PC2 is a dynamic, distributed real-time system designed to
manage and control Programming Contests.
It includes support for multi-site contests, heterogeneous platform
operations including mixed Windows and Unix in a single contest, and dynamic
real-time updates of contest status and standings to all sites. This manual describes the steps required to
install, configure, and run a contest using PC2. Further information on PC2,
including how to obtain a copy of the system, can be found at http://www.ecs.csus.edu/pc2.
PC2
operates using a client-server architecture.
Each site in a contest runs a single PC2 server,
and also runs multiple PC2 clients which communicate
with the site server. Logging into a
client using one of several different types of PC2 accounts
(Administrator, Team, Judge, or Scoreboard) enables that client to perform
common contest operations associated with the account type, such as contest
configuration and control (Administrator), submitting contestant programs (Team),
judging submissions (Judge), and maintaining the current contest standings
(Scoreboard).
PC2
clients communicate only with the server at their site, regardless of the
number of sites in the contest. In a
multi-site contest, site servers communicate not only with their own clients
but also with other site servers, in order to keep track of global contest
state. The following communication
requirements must therefore be met in order to run a contest using PC2: (1) a machine running a PC2 server
must be able to communicate via TCP/IP with every machine running a PC2
client at its site; and (2) in a
multi-site contest, every machine running a PC2 server must be able
to communicate via TCP/IP with the machines running PC2 servers at
every other site. In particular, there must not be any
'firewalls' which prohibit these communication paths; the system will not
operate if this communication is blocked.
It is not necessary for client machines to be able to contact machines at other
sites.
Each PC2 module (server or client) reads one or
more '.ini' initialization files when it starts; these files
are used to configure the module at startup.
The client module also tailors its configuration when a user (Team,
Judge, etc.) logs in. In a typical PC2
contest configuration, each Team, Judge, etc. uses a separate physical machine,
and each of these machines runs exactly one client module. It is possible to have multiple clients
running on the same physical machine, for example by having different users
logging in to different accounts on a
shared machine. In this case, each user
(Team, Judge, etc.) will be executing their own “Java Virtual Machine (JVM)”,
and must have their own separate directory structure – including their own
separate copy of the PC2 initialization files in their account.
Setting up and
running a contest using PC2 involves
the following steps: (1) installing Java
and PC2 on the contest machines; (2) creating/editing the necessary
initialization files; (3) starting the server(s) and clients(s); (4)
configuring PC2 for the contest via an Administrator client; and (5) starting the contest so that users
(Teams and Judges) can log in. These
steps are listed in checklist form in the next chapter, and are described in
detail in the remainder of this manual.
For those people
who hate to read manuals and would rather take a chance with a shortcut list,
here is a very terse summary of the
steps necessary to install PC2 and get it running. Please note that this is provided as a
convenience for geniuses (or gluttons for punishment). The remainder of the manual was written to
help everyone else. If you have problems
after following this list, please read the rest of the manual before
sending us a request for help.
In the event that the preceding checklist is a bit too terse, the
remainder of this manual discusses the details of using PC2 to
configure and run a contest. The first
step is to install the necessary software, as described in this chapter. The remaining chapters of the manual cover
initialization files, starting the system, configuring the system for a
contest, starting the contest, using the PC2 scoreboard, and special
considerations for multi-site contests without an active network connection
between sites. In addition several
appendices cover details of certain topics.
1. Install
the Java Software Development Kit (SDK) or Java Runtime Environment (JRE), version 1.3.1 or later on each machine. The remainder
of this manual assumes that “$JAVAHOME” represents the SDK installation
directory. For information on obtaining
the Java SDK, visit http://www.javasoft.com.
2. Add
“$JAVAHOME/bin” directory to the PATH environment variable on each machine
(i.e., for each user).
3. Create a
directory on each machine to house the PC2 system. The remainder of this manual assumes that
“$PC2HOME” represents the PC2 installation directory.
4. Add “.”
(‘the current directory’), “$JAVAHOME/lib”, and “$PC2HOME” to the CLASSPATH
environment variable on each machine.
(Note: the JRE does not use the CLASSPATH variable; when using the JRE
instead of SDK it is necessary to use the “-cp” parameter. See the JRE documentation for details.)
5. Download
the file pc2v85.zip (or
gzip) from the PC2 website
into the $PC2HOME directory (that is, into the directory into which you want PC2
installed).
6. Unzip the
pc2v85.zip file,
being sure to tell the unzip program to “retain directory hierarchy” and
“preserve case sensitivity”. This should
create directories named “pc2”, housing the PC2 system, and “com”,
housing the IBM common Java facilities.
It should also create several “.ini”, “.bat”, and other files, along
with directories named “doc” and “samps”,
which contain the system documentation and a variety of sample scripts,
files, and other goodies you might want to examine. The file “index.html” can be used to browse
the documentation.
7. Add the
“$PC2HOME” directory to the PATH environment variable on each machine (i.e.,
for each user).
To uninstall PC2 it is only necessary to
undo the above installation steps; that is, remove the $PC2HOME directory and
its contents and restore the system environment variables to their former
values . PC2 itself does not
make any changes to any machine locations outside those listed above either
during installation or execution. In
particular, for example, it makes no entries in the “registry” in a Windows
environment, nor does it copy any files
to locations outside the installation directories in any environment.
When a
PC2 module (server or client) begins running, it reads a set of one
or more “initialization files” from the directory in which it was started. The Contest Administrator must ensure these
files are present as needed, and edit them as necessary, prior to starting a PC2
module. This chapter describes the
principle initialization files and their contents (note: some default versions of the
initialization files are provided with the PC2 distribution package;
these must be edited as necessary).
Further descriptions of initialization files and their contents can be
found in the Appendices.
This file is used to tell servers the names associated with sites in the
contest. It must be present in the
startup directory for each server (recall that there is one server per
site). Each line in the file gives the
name of one site in the contest. Site
names are case-sensitive.
Edit the file “sitelist.ini” in
the $PC2HOME directory on each machine.
Delete the default information in the file if it is there, and replace
it with one line for each site in the
contest, giving each Site Name (one site per line).
Ř
NOTE: it is VERY IMPORTANT that ALL SITES IN THE
CONTEST HAVE IDENTICAL sitelist.ini FILE CONTENTS -- i.e. the sites must be listed in the same
order in the sitelist.ini
file at all sites, and the spelling and
spacing of site names, including any embedded blanks, must be the same
in all sitelist.ini
files.
Ř
To avoid errors due to mismatched sitelist.ini
files, it might be desirable to have the Head Judge or Contest Administrator
create one sitelist.ini
file and transmit identical copies to each of the separate Contest Sites
(taking into account different line terminators for different OS environments).
Every PC2
module reads a file named “pc2v8.ini”
at startup. This file provides key
initialization information to the module.
The file is formatted in sections.
Each section starts with a section-name in square brackets; for example,
[server'. Each PC2 module reads the entire
file, but silently ignores any information which does not pertain to it. All lines starting with “#” or “;” are
comments and are also ignored, as are blank lines. The following different section names are
recognized: [server', [client' , [admin', [judge', [team',
and [board'.
Each section is made up of
lines containing "attribute assignment" statements of the form
attributeName=value
The
“attributeName” is a predefined string chosen from
list of PC2 configuration attributes. The “value”
is the value to which the corresponding attribute is set when the pc2v8.ini file is read by
a module. No spaces are allowed in front
of the “value” after the equal-sign.
Some
attribute assignment statements are specific to particular sections and have no
meaning for other sections (or for the modules that read them). Other attribute assignments are relevant to
multiple sections/modules and can appear in different sections.
A
complete list of the predefined attributes is given in the Appendices. Most attribute assignments are optional and
default values are used if an attribute assignment is not present in the pc2v8.ini
file. However, certain attributes must
be specified in the pc2v8.ini
file in order for the system to function properly.
Specifically,
on a server machine the [server' section must
contain an attribute assignment giving to the server the name of the site it is
serving, using a name specified in the sitelist.ini file. Also, the [client'
section on a client machine (Administrator, Team, Judge, or Scoreboard) must
contain an attribute assignment giving the name of the site (from the sitelist.ini
file) to which it belongs, and also must contain an attribute assignment giving
the IP address and port number at which it should contact its server. (Note:
PC2 servers expect by default to be contacted initially on
port 50002; see the appendices for further details on port number usage in PC2.)
Based
on these rules, a minimum sample pc2v8.ini file is shown below. Note that since all modules ignore sections
and attributes which do not apply to them, it is permissible to create a single
pc2v8.ini
file containing all the required entries and put this same file on all machines
at a given site.
# sample pc2v8.ini file for
site named “Site1”
[client'
# tell the client what site
it belongs to
# and where to find its
server (IP and port)
site=Site1
server=192.168.1.117:50002
[server'
# tell the server which site
it is serving
site=Site1
The
sample pc2v8.ini
file shown above would be appropriate for both client and server machines in a
single-site contest (except that the IP address in the [client' section should
be replaced by the actual IP address of the site server, and the site name
should be changed if necessary to match the site name given in the sitelist.ini file). Alternatively, a file containing just the [client' section shown in the sample could be put on all the
client machines, and a different file containing just the [server' section
shown in the sample could be put on the server machine.
For a multi-site contest, one additional requirement exists for the pc2v8.ini
file. In a multi-site contest, the
server at one of the sites is designated the “primary server” and servers at
all other sites are designated “secondary servers”. When a primary server is started, it waits
for other servers to contact it. When a
secondary server is started, it automatically attempts to contact the primary
server; this is how the inter-server communication in a multi-site contest is
established. (The distinction between
whether a server waits to be contacted (is a primary) or initiates remote
contact (is a secondary) is in fact the only distinction between “primary” and
“secondary” servers.)
In a multi-site contest exactly one of the servers should be started as a
primary server (it does not matter which site has the primary server; once
communication is established all sites run as peers). The servers at all other sites should be
started as secondary servers.
Designation of a server as primary or secondary is controlled by the
contents of the pc2v8.ini
file.
By default (that is, in the absence of any information in the pc2v8.ini to the
contrary), a server assumes it is a primary server when it starts. Designating a server as a secondary server is
accomplished by providing an additional entry in the [server'
section of the pc2v8.ini
file of the server. This additional
entry, known as the remoteServer
attribute, tells the secondary server the IP address and port number at
which it should attempt to contact the primary server. If this remoteServer attribute is not
present in the pc2v8.ini
file when a server starts, the server implicitly assumes it is a primary
server.
Thus for example, the pc2v8.ini file on
the primary server in a multi-site server might look like the entry in the
sample above (since it does not contain any remoteServer attribute), whereas the pc2v8.ini file on
machines at a second site might look like:
# sample pc2v8.ini for a
second site
[client'
# tell the client what site
it belongs to
# and where to find its
server (IP and port)
site=Site2
server=192.168.1.104:50002
[server'
# tell the server which site
it is serving
# and how to contact the
primary server
site=Site2
remoteServer=192.168.1.117:50002
Note
that the (sample) IP address given in the [client' section of this pc2v8.ini file
for Site2 (a secondary site) is the (hypothetical) IP address of the server for that site,
whereas the IP address given in the remoteServer attribute in the [server' section is the address of the primary server – the
address which the (secondary) server at Site2 should use to contact the primary
server and “join the contest”.
When a server at a site is started, it reads a file name “reject.ini” which
contains the “reject messages” which can be sent from the Judges back to Teams
when the Team submits a program which is not a correct solution. The Contest Administrator must edit the file reject.ini in the
$PC2HOME directory on the server machine so that it contains one line of text
for each “NO” message to be sent from the Judges to Teams when a run is
rejected.
For example, if the Contest rules specify that the Judges will return one
of the three messages “Runtime Error”, “Time Limit Exceeded”, or “Wrong Answer”
for failed runs, put each of these messages on a separate line in the reject.ini file. Note that the reject.ini file must exist and be
properly edited prior to starting the server at a given site.
The contents of the reject.ini file thus determines
the set of choices from which a Judge can choose when responding to a team’s
submission. PC2 automatically
provides the Judge with a choice of “Yes” (this is a correct solution to the problem),
plus each of the messages listed in the reject.ini file.
PC2 automatically prepends
the characters “No –
” to each “reject” message sent to a team when a Judge determines that the
submission is not a correct solution to the problem. The text strings in the reject.ini file should contain
only the distinguishing portion of the message.
The Contest
Administrator should insure that every site in a multi-site contest has the
same set of reject message strings in the reject.ini file at the site.
Once PC2 has
been installed and the necessary “.ini” files have been properly set up (edited), the
normal PC2 startup procedure is to start a primary server, then
start an Admin client connected to that server and use the Admin client to
configure the contest details (problems, languages, etc.). Once the contest has been fully configured
using the Admin client, secondary servers can be started at remote sites (if
any), followed by additional clients at both the primary and (in the case of a
multi-site contest) secondary sites.
The PC2
distribution comes with a collection of “command scripts” designed to simplify
starting the various modules. The
available command scripts and their corresponding functions are listed below. To invoke the specified function, simply type
the corresponding command at a command prompt while in the $PC2HOME directory.
|
Command
|
Function
|
|
pc2reset
|
Creates a backup zipped archive of all existing PC2
databases, then clears all databases to the state matching a fresh
installation.
|
|
pc2server
|
Starts a PC2 Server
|
|
pc2admin
|
Starts a PC2 Client expecting an Administrator
login
|
|
pc2team
|
Starts a PC2 Client expecting a Team
login
|
|
pc2judge
|
Starts a PC2 Client expecting a Judge login
|
|
pc2board
|
Starts a PC2 Client expecting a Scoreboard
login
|
Thus the normal startup
procedure
would be to type the command “pc2server” to start a
server, then in a separate command window to type the command “pc2admin” to start an Administrative client to be used to
configure the contest details in PC2. Once the contest details are configured,
other clients can be started (as well as servers at other sites) using the
appropriate commands.
When a server is started
(using the command “pc2server”), the user will be prompted
with a question, as follows:
Are
there already servers up (Y, N, or Q) ?
The server asks this question to help it
determine whether this is a normal contest startup, or whether instead this is
a server which is being “restarted” following a crash. The following describes how the question
should be answered.
For a single-site contest, the
answer should always be “N”, since there should never be any other servers
running (the Contest Administrator must ensure that any previously-running PC2
servers – e.g. from a practice session – have been shut down prior to starting
a server).
For a multi-site contest:
Ř
If this
is the first startup of the primary server, the answer should be “N” since
there should not be any other servers running.
In this case the server will simply wait to be contacted by other
(secondary) servers.
Ř
If this
is the first startup of a secondary server, the answer should be “Y” since the
primary server should already have been started. In this case the (secondary) server will use
its pc2v8.ini information to contact the primary server.
Ř
If this
is a restart of a server which crashed or was shut
down for some reason during a contest (thus this is a case of a server
attempting to rejoin a contest which is already in progress), the answer
should be “Y” since there should still be other site servers running. In this case additional considerations
apply; refer to the Appendix “Restarting a Server” for further details.
Once a server has been
started according to the preceding steps,
the server displays a series of log messages on the screen. If the
pc2v8.ini file indicates
this is a secondary server (i.e., it has a “remoteServer=xyz” entry) the server will attempt to connect to
the primary server (as specified by “xyz”).
In either case, when the server is fully initialized it will display a
message similar to:
Fri May 09 19:34:52
PDT 2003 main Server at site "xxxxxxx" is ready. (id=1)
where
“xxxxxxx” is the name of the site as it appears in the sitelist.ini file. If any
errors occur or the server fails to produce the “server is ready” message,
check the files “pc2.log” and “pc2serv.log” for further details.
Once a server is
running it can be stopped simply by terminating the Java VM under which it is
running. This is typically done, for
example, by typing ^C (control-C) at the command window from which the server
was started. Note that stopping a server
will immediately terminate all PC2
activities at that site.
Once a PC2
server is running at a site, users (Contest Administrators, Judges, and Teams) at the site can start PC2
clients to login and use the
system. The normal procedure is first to start a
client using the “pc2admin”
command and login as the “root” administrator in order to configure the
contest. Subsequently each Contest Judge
would start a client using the “pc2judge” command, and each Team would start a client
using the “pc2team”
command. The Contest Administrator would
normally also start a PC2 scoreboard using the “pc2board”
command, logging in using the PC2 account “board1”.
Each time a
client is started, the client will read its pc2v8.ini file to determine its site name and the location of
its server, and then contact the server.
Following this initialization sequence, the client will display a
“login” window as shown below, indicating that it is ready to accept a user
(Team, Judge, Administrator, or Scoreboard) login. Depending on the logging levels specified in
the client’s pc2v8.ini file,
the progress of these steps will be displayed and/or written to the file pc2.log in the client’s startup
directory. If any errors occur or the
client fails to produce the login screen, check the “pc2.log” file for details.
Once PC2
is set up and running, it is necessary to “log in” to the system via a
client login window in order to use the system. A PC2 “account” is required in order to log
in. Initially only a single account exists; the
account name (“login ID”) of this single account is “root”. This account is a master
Administrative account which is used to configure the PC2 system initially for
the contest. Regular users (especially Teams) should NOT be given access to
this account.
The
default password for the root
account is root (same as the login name). Note that the default master password is
given right here in this paragraph of this document, which is publicly
available on the Web.
Caveat Administrator : change the root password!
Passwords
can be changed via the “Manage Accounts” function on the “Accounts” tab; see
below.
After logging in to the “root”
Administrator account, the following screen will be displayed. This is referred to as the “main
Administrator screen”. It provides a
series of “tabs” across the top to select various contest administration
functions. The tabs which are used to
configure the contest prior to starting are described in the remainder of this
chapter. Tabs used to start the contest
and monitor its progress are covered in the following chapters.
Before
any logins other than root can
occur, it is necessary to create user accounts.
To create accounts for users, click the Generate button on the Accounts tab on the main Administrator
screen. This will display the following
screen:
Note that the (1) to the right of the
“Administrators” label means that currently one Administrator account exists –
that one account is the root account
– and no other accounts exist. It is
necessary to create one Team account for each team at this site, one Judge
account for each person who will be judging the contest at this site, and at
least one Scoreboard (“board”) account if the PC2 scoreboard is going to be used at this site for tracking
contest results. (It does not hurt to
generate a few extra accounts in each category, for flexibility.)
Enter the desired number in each
box, then click the Generate
button. Depending on the
number of accounts, number of sites, and communication delays, it will take anywhere from a few seconds to
several minutes for the account generation to complete. Once the accounts are generated the system
will automatically return to the main Administrator screen. PC2 accounts always start with the
word team, judge, admin, or board, followed by a number. See the next section for information on
viewing/changing the state of generated accounts.
Accounts in PC2 are site-specific. This means that in a multi-site contest, an
Administrator at each site must login as root and create the accounts
for that site. (It is possible to have a
single person handle account creation for all sites in a multi-site contest, by
starting separate Administrator clients which have server IP addresses in their
pc2v8.ini files pointing to different site
servers. In any case, however, it is
necessary to start a separate Administrator client connected to each
site server and create the accounts to be used at that site; there is no way in
the current version of PC2 to perform multi-site account creation
using just a single Administrator client.)
Each generated account will be created with a
password, but at present the default (only available) specification for
passwords on newly-generated accounts is “Passwords same as Account Name”. This means for example that the password for
the account “team1” is “team1”; the password for the account “judge1” is
“judge1”; etc. Each account name and
password are created with all letters in lowercase.
Passwords for accounts can be changed from
their default values by editing each account.
To do this, click on the “Manage Accounts” radio button on the Accounts tab of the main Administrator screen, then use the “Account Type” drop-down list to select
the type of account you wish to manage (change). For example, the following screen displays
the list of Team accounts.
To edit an account, click on the account in the display grid to select
it (“team2” has been selected in the display shown above), and then click the
“Edit” button. This will display a new “Edit Account” window, shown below:
The “Display Name” for an account is the
name which will appear on the PC2 Scoreboard; this can be set to any
desired value (such as the name of the team’s school, or the team member’s
names). The “Password” and “Verify
Password” fields can be used to set any desired password for the account.
The “Active” checkbox determines whether a
team account will be considered in computing the scoreboard standings; if there
are some team accounts which will not be used then you should uncheck their
“Active” checkboxes – otherwise they will appear on the scoreboard as teams
which have solved no problems. Note that the “Active” checkbox only
determines whether a team appears on the scoreboard; teams which are not
“Active” can still log in, submit runs, and otherwise participate in the
contest. This is designed to allow
“guest” or other “non-competitive” teams to participate. (To prohibit any activity from a team
account, change the account password.)
The “Region ID” field is used to associate accounts with different
“regions” or “groups”. This is used in
conjunction with the PC2
scoreboard for displaying rankings of different subgroups (see the chapter on
the PC2 Scoreboard
for further details).
Note that PC2 accounts are unrelated to any user accounts
which may otherwise exist on the systems being used for the contest (for
example, user accounts provided by the operating system).
In a multi-site contest, newly created PC2 accounts are
automatically distributed throughout the entire system, including across
multiple remote sites. As previously
noted, accounts are “site-specific”.
Note also, however, that accounts at different sites are numbered using
the same sequence; the first team
account at Site 1 is called “team1”, and the first team account at Site 2 is also
called “team1”, etc.. Accounts are
therefore identified by always giving both the Site number and the Team number,
as in “Site1Team1”, which is a different account from “Site2Team1”.
Since editing account data (e.g. Display Names, Passwords, etc.)
interactively for every account is cumbersome, it is desirable to be able to
prepare the data “offline” ahead of time and then load it into PC2. This can be done by preparing an “account
data” file and using the “LOAD” button on the “Manage Accounts” screen to load
the data into the system.
An “account data load file” consists of a series of text lines, one for
each account to be configured. If it is
desired to load data for different types of accounts (teams and judges, for
example), it is necessary to prepare a separate account data file for each type of account. This is necessary because the LOAD button
applies the loaded data to the “currently selected” type of account as shown in
the “Account Type” dropdown list on the
“Manage Accounts” screen.
The format of the account data load file is as follows. File lines starting with ! or # in the first
column are ignored. Each non-comment
line has FOUR fields, separated by a pipe (|). The fields are:
Account_number
Display_name
Active (either "true" or "false")
Password
For example, to initialize accounts
“team1”, “team2”, and “team3” so that the “team1” account displays on the
scoreboard with the name “Number 1” and has a password of “pass1”, while the
“team2” account displays on the scoreboard with the name “Team Number 2” and
has a password of “myPass”, and the “team3” account is made inactive (does not
display on the scoreboard), the following entries would be placed in the load
accounts data file for teams:
1|Number 1|true|pass1
2|Team Number 2|true|myPass
3|My School Name|false|
Empty (or
blank) Display Name and/or password fields cause the corresponding item to be
set to the empty string. An empty (or
blank) "active" field is considered false. Imported values overwrite any values that
were in the system previously. Also, it
is not necessary to provide a record in the data file for every account; the
Account Number field determines which accounts will be modified (all unlisted
accounts will remain unchanged).
Note that
there is no way in the current version of PC2 to assign a “Region
ID” to an account via a “Load” operation; if regional groupings are being used
then the Region ID values must be assigned manually (see the chapter on the PC2
scoreboard for further details).
PC2 was designed for supporting the ACM International
Collegiate Programming Contest, including its local and Regional contests
worldwide. The ICPC maintains an online
Contest Registration system which is used by Regional Contest Directors (RCDs)
around the world to manage participation in the various ICPC Regional Contests. PC2 provides interfaces to import
contest registration data from the ICPC Registration system, and also to export
contest results back to the ICPC web site.
See the Appendix on ICPC Import/Export Interfaces for further
information on importing/exporting ICPC Registration system contest data.
During a contest it is possible (likely, it seems from experience) that a
team or judge will somehow manage to kill their PC2 Java VM after
having logged in to PC2. This
can happen in a variety of ways, such as a total machine crash or a simple
inadvertent command to “kill” the “command window” which is running the Java
VM (users should be warned not to terminate
this command window).
In the event that a user becomes disconnected and/or kills PC2
somehow, they will (eventually) wish to log back in to the system. However, in the current version of PC2
the server has no way of detecting the loss of connectivity to the client
machine; the server therefore will think
they are still “logged in”. For security (to prevent teams from logging
in and acting as another team) PC2 refuses to allow multiple logins
of the same account. Thus a disconnected
user may not be able to log back in (because the server thinks they are
“already logged in”).
As a compromise in this situation, PC2 uses the following rule
when it receives a login request from a client:
if there is already a registered client login using that account,
then if the IP address of the machine requesting the new login is
identical to the IP address of the currently registered login for that account,
then the server attempts to communicate with the already-registered client (in
a sense it ‘pings’ the client). If this communication fails, PC2 assumes this is a case of a client which
became “disconnected” somehow, and it discards the old login
registration and accepts the new registration instead. This procedure effectively means that most
“disconnections” can be resolved simply by having the team (or judge) log back
in (this was not true in earlier versions of PC2).
If for some reason a client was logged in, got disconnected, and had to
change to a machine with a different IP address, the server will not recognize
this as a reconnection login, and it will enforce the rule prohibiting multiple
logins of the same account (since it will still have the old login
record). To solve this, the Accounts tab on the main Administrator screen provides
a function to “force a logout” of a currently logged-in account.
To invoke the “forced logout” function, click on the “View Logins” button on
the Accounts tab, then click on the row in the display
showing the user account that needs to be “logged out”. This will activate the “Logout” button (see the Accounts tab
screen, above). Pressing this button
will then force the Server to drop the login connection, allowing that account
to re-login, whether from the same IP address or not.
PC2 must be provided with information
about the problem set to be used in the contest. To enter this information, click on the Problems tab at the top of the main
Administrator screen. This will produce
a display similar to the following:
Note that
initially no problems are listed since none have been added to the system. To add a problem, click the Add button. This will
produce the “Edit Problem”
dialog shown below.
To define a
contest problem to the system, perform the following steps using the Edit Problem dialog:
Ř
Enter the problem name in the top textbox.
Ř
If the problem requires an input data set, click
the “Problem Requires Input Data” checkbox and then
1.
select either “Stdin” or “File”, depending on whether
the problem description tells teams to write their programs to obtain input
data from “standard input” or from a file,
then
2.
use the Browse button to select the data file.
Ř
If the Judges have provided an “Answer File” (a
file showing the expected output of a program correctly solving this problem),
click the “Judges Have Provided an Answer File” button and then use the Browse
button to select the Answer File.
Ř
Click the Update
button to store the problem information.
As
each set of contest problem information is entered, it will be displayed on the
main Administrator screen (when the Problems tab
has been selected). To change some previously entered information for a
problem, click on the problem row in the main display to select it, then click
the Edit button.
This will return to the Edit
Problem dialog, where changes
can be made.
The following additional notes apply when entering data using the Edit Problem dialog:
Ř
The Run Timeout Limit value (shown as 120 in the
sample screen above) is not enforced by PC2, in the sense that the
system does not automatically stop a submission from executing when the time
limit is reached. Rather, a count-up
timer is displayed during program execution so that the Judge can tell how long
the program has been executing, and the counter value turns RED when the specified
timeout limit is reached. The timer also
includes a button to allow the Judge to terminate the program
Ř
The content of the
input data file for a problem is stored internally when the Update button is pressed (that is, PC2 makes an internal copy of the file). For this reason, editing the user’s copy of the file will not automatically
change the data presented to team programs.
To modify the data file for a problem, the contest administrator must
EDIT THE PROBLEM and load the new copy of the data file, then press the Update
button again.
Ř
The Validator tab on the Edit Problem dialog is used to interface an automated judging program for this problem
to PC2. See the
Appendix on Validators for further details.
Ř
All team program output is expected to go to
“standard output” (where it is captured by PC2 and saved for display
to the Judges). More specifically, there
is no mechanism in the current version of PC2
for dealing with programs which are written to send their output to a
destination other than “stdout” (for example, programs which send their output
to a file).
Ř
Contest problems in PC2 are global,
in the sense that once a problem definition is entered by the Contest
Administrator that problem definition is broadcast to all sites. There is no mechanism for having teams at
different sites in a multi-site contest see different descriptions of the same
problem. If there is some reason that
different descriptions are needed for the same problem (for example, if a
problem needs to be described differently at different sites due to OS
differences), it is necessary to enter the different problem descriptions
effectively as different problems.
(While this is not very elegant, it is also something that we rarely –
virtually never – see in real contests…
Said another way, PC2 views a contest as a set of teams all
working on an identical problem set.)
Note also that contest problems must appear in the same order at
each site in a multi-site contest; see the chapter on Loosely-Coupled
Multi-Site contests.
Ř
PC2 copies the input data file for a
problem into memory each time a team program for that problem is executed (this
is how the architecture manages the insertion of the input data into the input
stream of the team program). If the size
of the input data set (file) for a problem is particularly large and the system
has a relatively small amount of memory, it is possible to exceed the memory
limits of the Java Virtual Machine (JVM).
This problem can be circumvented by utilizing a script for the “Program
Execution Command Line” which copies the required data file into the $PC2HOME/execute
directory and then invokes the team program.
(See the following section on Contest Languages, and the Appendix on
Language Definitions, for further details.)
PC2 must be provided with information about the programming
languages used by contestants (Teams).
To enter this information, click the Languages tab
on main Administrator screen. This will bring up a display similar to the
following:
Note
that the display is empty because no languages have been defined yet. To add a language description, click the Add button. This will bring up an Edit Language dialog, similar to the one shown below, containing
four fields used to describe the language to PC2.
The “Display Name” for a language is the name which Teams will see when
they are asked to specify the language in which they have written a program
which they are submitting. The Display Name can be any arbitrary text; it does
not have to be a real language name (for example, “Local C Compiler” could be a
legitimate language Display Name).
The “Compile Cmd Line” field is
used to specify the command line which is used to compile source code and
produce an “executable program file” in the language.
The “Executable Filename” field is used to tell PC2 the name
(or more correctly, the form of the name) of the output (executable program)
file produced by the compilation process.
PC2 clears its internal execution directory of any instance
of the specified executable file prior to compilation, and checks for the
existence of the specified executable file following compilation. It interprets the existence of a new
executable file as evidence of successful compilation.
The “Program Execution Command Line” field is used to specify the (form
of the) command line required to execute (run) the resulting program. Execution is only performed if the preceding
steps were successful in producing a new executable file.
The Contest Administrator must define each language to be used in the
contest by filling in the four language definition fields, replacing the
default values shown above with the appropriate values for the language being
defined. (Note that the preceding example
screen shows values called “command parameter substitutions” in the language
definition fields; see the following sections for further details on the
definition fields.)
Once the definitions for a language have been entered, click the Update button
to store the information and return to the main Administrator screen. The language names will be displayed under
the Languages tab on the main Administrator screen, as shown
below. To add more languages, click the Add button again to return to the Edit Languages screen. To modify a previously-entered
language, click on the row containing the language description to select it and
then click the Edit button.
See the Appendix on Language Definitions for further details.
The four language description fields in the Edit Language dialog can be “hard-coded” by
entering fixed values if desired. For
example, the Display Name for a language is normally fixed for the duration of
a contest (e.g., “Java”, or “C++”, or “Pascal”).
However, entering fixed values for the Compile Command, Executable
Filename, and Program Execution Command fields can be extremely cumbersome and
inflexible – the details of these fields may need to change with each different
program file submission, for example. In
order to provide more flexibility, PC2 supports the use of
“parameter substitutions” in these fields.
PC2 parameter substitution fields are indicated by matching
curly braces, with the first character inside the left curly brace being a
colon (‘:’). Following the colon
character is exactly one of the predefined PC2 parameter
substitution keywords. Any number of command parameter substitution fields may
appear anywhere in a language description field. The currently defined
parameter substitution keywords and their corresponding meanings are given
below.
|
Keyword
|
Meaning
|
|
|
Replace with the full name of
the submitted file, including any extension (but excluding any ‘path’
specifier on the front of the filename)
|
|
basename
|
Replace with the base component
of the file name, omitting any extension (and excluding any ‘path’ specifier
on the front of the filename)
|
The following section shows examples of language definitions, including
the use of command parameter substitution fields.
The following screen shows a set of filled-in fields defining a language
named “C” and using the GNU GCC compiler.
The compile command line invokes the compiler (“gcc”) and passes it an
argument specifying use of the math library (“-lm”). The compile command line also specifies the
assignment of a specific name to the “object” (compiled output) file (the “-o”
argument), followed by the name to be assigned to the object output file. In this case, the object output file is to
have the same name as the base name of the input source code file. (So for example if a team submitted a file
named “proga.c”, the object output file would be named simply “proga”, since
that is the value to which the “{:basename}” substitution parameter would be
expanded.)
The final argument on the compile command line gives the name of the
source file to be compiled, which would be expanded from “{:mainfile}” to
become “proga.c” if that was the name of the submitted main program source
file.
The Executable Filename field indicates that the executable file which
produced by the compile command has the same name as the base name of the
submitted program; this is because the compile command specifies (via the “-o”
argument) that this is the executable file name which should be produced.
The Program Execution command field specifies that the command used to
execute the compiled program is simply the same as the name of the executable
file produced by the compilation step (and specified in the Executable Filename
field), which in this case is again the base name of the original source code
file.
If a team were to submit a C program in a file named proga.c using the above language, PC2
would first execute:
gcc –lm –o proga proga.c
to
compile the program (substituting proga for the {:basename} parameter
and proga.c for the {:mainfile}
parameter). It would then check for the
existence of an executable file named proga, and if that file exists then PC2 would
request the underlying operating system to execute the command:
proga
The following screen shows
a second language definition example: a
definition for a language with the display name “java”:
If
a team were to submit a Java program in a file named sumit.java using this language
definition, then PC2 would
execute the following command to compile the program:
javac sumit.java
Then PC2 would check for
the existence of an executable file named
sumit.class and if that file exists then PC2 would
execute the following command to run the program:
java
sumit
Note
that the form of the language definition fields differs somewhat between the
previous example (C) and the second example (Java). This is because of the different ways in
which these two languages define the compilation and execution process. Notice also, however, that while the language
paradigms are different, the use of
command parameter substitutions allows the Contest Administrator easily
to provide descriptions of how to handle the differences. The appendices contain further samples of
language definitions for specific compilers.
Language definitions in PC2 are global. This means that, just as with Contest Problem
definitions, when a language definition is entered at one site in a multi-site
contest, that language definition will be visible at all connected
contest sites. However, unlike the
situation with Contest Problems (where the problem definitions are usually
identical across sites), language definitions may differ between sites – even
for the “same language”.
For example, it may be the case that every site allows the use of the “C”
language. However, it may also be true
that the specific command sequence to invoke the C compiler may differ between
sites: a different C compiler might be
used at different sites, or even if the same compiler is used it may be
necessary to allow for differences in the “path” needed to access the compiler
or for other environmental differences.
One way to deal with differences in language details between sites is to
create a different PC2 language description for each different
language/site combination. This can
quickly become cumbersome, however; for
example, if there are four languages (e.g. C, Java, Pascal, and Perl) and five
sites using those languages, it could require entry of up to 20 different
language descriptions (Site1C, Site2C… Site1Java, Site2Java,… etc.). This can become particularly unwieldy for Teams, who must search through a list of
20 different languages looking for not just the correct language but the
correct language for their site.
To avoid this combinatorial explosion of language definitions, a simple
technique can be used when defining languages in a multi-site contest: use of generic language scripts,
tailored at each site for the site-specific configuration.
For example, consider a contest using, say, C, Java, and Pascal. The Contest Administrator should define those
three languages in PC2 using the actual language names (“C”, “Java”,
and “Pascal”) as the PC2 “language Display Names”. However, rather than defining a specific
compilation command for each language (which may differ between sites), each
language should have as its compilation command a command which invokes a
language-specific (but site-independent)
script (or “batch file”) designed to compile a program in that
language.
In other words, for the above three languages, PC2 language
definitions would be created to define the “compilation command” for the
language named “C” to be the invocation of a script (batch file) named “compileC”
(or “compileC.bat”); the compilation command for Java would be the
invocation of a script named “compileJava”; and the compilation command for Pascal would be the
invocation of a script named “compilePascal”.
Then, at each site, the
Site Director is responsible for placing on machines at that site a set of
scripts or batch files of the corresponding names (e.g. compileC,
compileJava, and compilePascal). Within
each script at each site is a set of site-specific commands which perform
the necessary steps (compile a C program, compile a Java program, or compile a
Pascal program) in the appropriate site-specific manner.
Note
that if necessary, the same technique of “generic scripts” which vary between
sites can also be used in specifying the details of “Program Execution Command
Line” for languages. That is, the Contest Administrator can specify “executeC”,
“executeJava”, and “executePacal” scripts for the program execution language
definitions in PC2 and then arrange for appropriately different
script contents at each site.
Note
also that PC2 “command parameter substitutions” may be used in
compilation and execution command lines independently of whether the command is
invoking a script or not; in this way the Contest Administrator can arrange to
pass necessary data (such as the main program file name and/or the base name)
to a script.
Using
generic script names in PC2 language definitions and providing
site-specific implementations of each language script at each site allows the
Contest Administrator to significantly
reduce the number of language definitions which teams must deal with, while at
the same time retaining the flexibility necessary for dealing with site
differences in a multi-site contest.
The Options tab
on the main
Administrator screen displays a screen
which allows selection of additional options which can be used to display and/or control
various aspects of a contest. These are
broken into two sets: General Options and Balloon Options; each of these categories
is listed and described below.
Selecting General Options displays a new screen with three tabs: Judge, Team, and Scoreboard. The
following tables list the option settings available on each of these screens.
|
Judge Options
|
Description
|
|
Show Team Numbers to Judges?
|
Specifies whether or not to
reveal the identify of Teams to the Judges while a run is being judged:
Yes – displays the team
name/number on each run
|
|
Team Options
|
Description
|
|
Maximum Output Size
|
Specifies the
maximum amount of output, in Kbytes,
which a team program is allowed produce to stdout or stderr. Any output beyond this amount is discarded
by the system. The default value is
512 (1/2 MB).
|
|
Disable Team’s Viewing
Judgements
|
Specifies the
number of elapsed minutes after which judgments are not displayed on team
grids. The default value if not
specified is 240 (4 hours).
|
|
Scoreboard Options
|
Description
|
|
Contest Title
|
Specifies the
contest title which is to be displayed on the scoreboard.
|
|
Submission Penalty Points
|
Specifies the
number of penalty points to be added to a team’s score for each “no” judgment
for a problem which is successfully solved.
The default value is 20 points.
|
In
many contests (including the ICPC World Finals), balloons are used to indicate
to contestants and spectators alike the general state of the contest. Each time a team solves a problem, a balloon
of a specific color is sent to the team and attached on or near their
machine. As the contest progresses, the
contest floor gradually fills up with a multi-colored display showing how
various teams are doing in the contest.
(This is a colorful and normally well-received operation; if you have
never tried it, we recommend doing so.)
Using
balloon notifications in a contest does present some additional management
overhead (keeping track of which team should get what color balloon,
etc.). Since PC2 was designed to support the
ICPC World Finals (as well as its Regional and Local contests), it contains
some built-in support for “balloon operations”.
Selecting the Options tab on the main
Administrator screen and then clicking on “Balloon Options” will
cause a dialog similar to the one shown below to appear; this dialog is used to
configure the handling of balloons in a contest.
Balloon options include ability to specify the color of balloon
associated with each problem; sending messages to a printer each time a balloon
should be delivered to a team; and sending an email message via a specified
email (SMTP) server to an arbitrary email account each time a balloon should be
delivered to a team. Printed and emailed
messages contain the relevant details such as Team, problem, and balloon color. Each of these options can be selected on a
per-site basis (so that, for example, sites can use different color balloons
for a given problem).
Generation of balloon
notifications is handled by the PC2
Scoreboard machine.
Once email and/or printing notification is enabled, every “YES” judgment
detected by the PC2
Scoreboard “board1” account will cause an email notification and/or a printed
notification to be sent to the configured location.
To enable the use of email balloon notifications, check the “Enable
Emailing” box, then enter in the
appropriate text boxes the full name of an SMTP email server accessible to the
PC2 Scoreboard
machine along with a valid email address.
To enable the use of printed balloon notifications, check the “Enable
Printing” box, then enter in the appropriate textbox the device identifier of a
printer accessible to the PC2
Scoreboard machine. If it is also desired to print notifications of “NO”
judgments, check the “Enable Sending No’s” box.
This tab on the Administrator main
screen displays a list of all the sites in the contest; this list should be
checked to verify that PC2 knows about all sites. If the site is currently active (connected
to the rest of the contest) the IP address for that site’s server is
displayed.
Once the contest has been has been fully
configured in PC2, the contest clock must be started before teams
can submit runs (the Team client will not allow run submissions if the contest
clock has not been started). The Time/Reset tab on the
main Administrator screen is used to control the contest clock display and to
start and stop the contest clock.
Clicking the Time/Reset
tab produces a screen similar to the one shown below:
The Start Contest Time button is used to tell PC2 to start the contest
clock for the current site (the current
site name is shown in parentheses in the title bar). Pressing Start Contest Time starts the contest clock
running for that site and allows teams
at that site to submit runs.
The amount of time
remaining in the contest at the current site is displayed in the “Remaining
Time” field, and automatically starts
counting down as soon as the Start
Contest Time button is pressed.
It continues to automatically update (count down) as long as the contest
clock is running.
It
is important to note that, from the point of view of PC2, the
contest does not start until the Start
Contest Time button is pressed.
For this reason it is important that the Contest Administrator
remember to press the Start Contest Time button
at the actual time the contest starts.
Failure to do this can produce erroneous scoring results.
Typically, for example, a
contest is deemed to have “started” when the contest problems are distributed
to the teams. If the PC2 Start Contest Time
button is not pressed for another, say, 15 minutes, then that 15 minutes will
not be considered to have been part of the contest by PC2. If a team were to submit a run 20 minutes
after the contest started (i.e. 20 minutes after the problems were handed out),
the timestamp on that run would show a contest elapsed time of 5 minutes, not
the correct value of 20 minutes. This
would produce erroneous values on the scoreboard.
The Stop Contest Time button is used to tell PC2 to stop the contest
clock for
the current site. The Stop Contest Time
button can be used to insert a pause in a contest (for example, to allow a
break for lunch). During the time the
contest is stopped, the contest clock at the site does not count down, and
teams are prohibited from submitting runs.
When the Start Contest
Time button is pushed again, the contest clock for the site picks up
where it left off.
Note that this means that
if a team submits a run one minute before the contest clock is stopped, and
then the clock is stopped for 30 minutes of real time, and then the team
submits another run immediately after the contest clock is restarted, the
timestamps on the runs will be one minute apart. In other words, PC2 does not
consider time during which the contest clock is stopped to be part of the
contest. (If this is undesirable – that
is, if the Contest Administrator wishes all time which elapses to be
counted, then simply do not press the “stop” button once the contest has been
started. )
PC2 can be instructed to automatically stop accepting runs
from teams at this site for scoring when the contest clock reaches zero (no
time remaining). This is done by
checking the “Stop This
Site Automatically” checkbox on the Time/Reset screen (the contest must currently be
stopped in order to change the state of the checkbox). If this checkbox is checked, then once the
contest is started any submissions from teams which are received after the
contest clock reaches 00:00:00 are automatically marked as “deleted” (refer to
the Editing Runs section,
below). Since runs marked as deleted
are not considered in scoring, late runs have no effect on the scoring of the
contest.
Even though late runs are not
considered in scoring computations, late runs are still captured by the
system. This allows the Contest
Administrator to make a decision later to accept such a run if desired; for
example, if it was determined that due to site-specific conditions the run should
have been accepted. Notifying the
system to accept a late run can be accomplished simply by using Edit Run to change the
run status to “not deleted”, along with setting the “Elapsed Time” (submission
time) of the run as appropriate.
When PC2 has been configured to automatically stop accepting
runs at the end of the contest and a run is received after the contest ends,
the team receives a notification back from the server informing them that the
run was received too late to be counted.
Also, even though late runs are marked as deleted and hence not scored,
they are automatically forwarded to the Judges. This allows the Judges to make a
determination of what the result of the run would have been if it had
been received within the time limit.
However, even if a Judge responds “YES” to a run, it still will not be
counted in the scoring if it was received after the clock reached zero and the
“Stop This Site
Automatically” box was checked.
If the “Stop This
Site Automatically” box is not checked, then PC2
continues to accept runs after the contest clock counts down past zero, and such
runs are treated (and scored) exactly as runs submitted before the clock
expired (except of course that their “submission time” will be larger). When the clock counts down past zero the
Remaining Time is displayed as a negative number, in RED. Thus the absolute value of the clock display
in this situation tells how long past the scheduled end of the contest the
clock has continued to run.
Note that “contest time”
in PC2 is site-based, and also that the “automatic stop”
feature applies only to the current site. These are intentional design features and are
important in order to allow sites in a multi-site contest to make adjustments
for site-specific conditions (such as the contest problems being handed out at
different times at different sites, or the need to allow for a pause at one
site). However, this means that contest
clock coordination in a multi-site contest can be critical; see the section
below on Multi-Site Clock Control if you are running a multi-site contest.
The Set Contest Times
button on the Time/Reset screen displays the Contest Time dialog (shown below) which allows
the Administrator to change the contest length to something other than the
default length (5 hours).
Entering a new time in the form HH:MM:SS in the “Contest Length” field
and then pressing the Set
button will set the overall contest length to the specified amount of
time. Setting a new contest length will
also automatically update the Remaining Time field.
Normally the contest length is set only once, prior to the start of a
contest, but this function can also be used to extend a contest after it has
been started. However, the contest
length can only be changed when the contest is stopped; the Set button will be
disabled when the contest clock is running.
Thus to extend a contest after it has started it is necessary to first
pause (stop) the contest, then change the Contest Length, then start the
contest again.
The time values in the Contest
Time dialog do not update automatically when the contest is
running; they display only the instantaneous time values at the moment the
dialog is activated. If the contest is
stopped when the dialog is activated, those times remain accurate indefinitely. If the Contest Time dialog is activated while the contest
is running, the times can be updated by pushing the Refresh Times button. (Recall that the time values cannot in any
case be changed unless the contest is stopped, so there is not usually any
reason for needing to view this dialog while the contest is running.)
If a new Contest Length which is less than the current Elapsed Time is
entered, the system displays a warning message to inform you that setting the
contest length to a value less than the elapsed time will effectively mean the
contest is over. If you decide to accept
this condition, then if “Stop
This Site Automatically” is checked then the contest will be
automatically stopped. If automatic stop
has not been enabled then the contest clock will indicate that time has moved
past the end of the contest by displaying the Remaining Time as a negative
number in RED.
The Reset Contest
button is used when it is desired to effectively restart the contest clock from
the initial conditions. It forces the
Elapsed Time to zero and the Remaining Time to be equal to the Contest
Length. Reset Contest is disabled when the contest clock
is running.
As described above, the
contest clock in PC2 operates on a per-site basis. That is, each site in a multi-site contest
has its own contest clock, and PC2
keeps track of “contest time” independently at each site. This is done to allow support for independent
time-management constraints at different sites, and allows scoring to be done
accurately without worrying about differences in timing between sites (e.g., a
necessary pause at one site which does
not affect other sites).
Each PC2 site
server determines the time of submission of a run from a team in terms of
“contest elapsed time”, which means that a submission will be marked
(“time-stamped”) according to the contest elapsed time at that site. The scoreboard in turn computes rankings
based on this “submission time”, which means that overall (multi-site) rankings
will be determined according to “contest elapsed time” at the site from which
each run originated. This method puts
teams at all sites on an equal competitive footing regardless of differences in
the time at which the contest actually starts at each site.
However, this mechanism
(keeping track of contest time independently at each site) can produce
erroneous scoring results if the Contest Administrator does not take care to
control the multi-site contest clocks correctly. Specifically, in a multi-site contest it
is important that the clock at each site be started at the moment the contest
starts at that site.
Typically, for example, a
contest is deemed to have started at a site at the moment the contest problems
are distributed to teams at that site.
If this event (problem distribution; contest start) happens at different
real times at different sites, then a Contest Administrator at each site should
press the Start Contest
Time button at that site precisely when the contest starts at
that site – regardless of whether the contest has started simultaneously at
other sites. In this way, runs submitted
by teams at each site will be correctly time-stamped with the true “contest
elapsed time” as their “submission time”.
If all sites in the
contest are fully connected, and the contest problems are handed out at the
same instant at all sites, then a single Contest Administrator can
easily coordinate the start of “contest time” at all sites correctly, simply by
using the “Start All Sites” button on the Time/Reset screen.
This button performs the same function as Start Contest Time except that it applies the
corresponding action (starting the contest clock running) simultaneously to all
connected sites. (Use the Admin Sites tab to determine
which sites are connected; connected
sites are those that have valid IP addresses displayed in the Sites screen).
In a fully-connected
multi-site contest where Contest Administrators at each site have agreed (e.g.
by telephone or other method) on the precise instant at which all teams at all
sites will received the contest problems and thus the time at which the contest
officially starts, having a single Contest Administrator press the Start All Sites button is the preferred (safest) way to
coordinate the start of a contest.
Likewise, if all sites are tightly coordinated, the Stop All Sites buttons can be used to stop the
contest clock simultaneously at all connected sites.
However, if there are some
sites which do not have network connectivity, or some sites at which the
distribution of the contest problems (and hence the start of the contest) is
delayed for some reason, it is critical that a Contest Administrator at that site makes certain that the contest clock is started at that site at the moment the problems are handed out (or whatever
other criteria determine the moment in time when the contest starts at that
site).
We
have seen more than one instance of a situation in a multi-site contest where
the contest problems were handed out at all sites (hence, the contest is
effectively under way at all sites), but one or more sites failed to notify PC2 that the contest had started
(either because the sites were not networked, or because the Start All Sites button was not used). In this case, if say 30 minutes elapsed
before PC2 is
notified to start its contest clock at one site, then teams at that site will
effectively get 30 “free” minutes – a run submitted 31 minutes after the
problems were handed out will appear to PC2 at that site to have been submitted “1 minute into the contest”.
Caveat Administrator:
It is critical that the PC2 “contest clock” be started, at every site, at the time
the contest starts at that site.
In addition,
recall that as noted previously the Stop This
Site Automatically checkbox is entirely local in function.
This box must be separately checked on the Admin client for each site
where it is desired that the contest stop automatically when the Remaining Time
clock counts down to zero.
In many contests, the overall contest activity starts with a “practice
session” prior to the start of the actual contest. The primary objective of the practice session
is to ensure that all teams are familiar with the operation of the contest
environment (PC2 in the present case) prior to the start of the real
contest. The practice session might
provide teams with a trivial “practice problem” to solve (“print your team
name” or “read a file containing integers and print the sum of the integers”,
for example), and then require all teams to login to PC2 and test
out the run submission mechanism by writing and submitting a solution to the
practice problem. Some contests also
require teams to practice using the PC2 “clarification system”
during the practice contest. A practice
session also has the advantage of giving the Judges practice with how PC2
works prior to the start of the real contest.
In order to run such a “practice contest” prior to the start of the real
contest, it is necessary to configure PC2 for the practice
contest. Most of the configuration is
identical to what is required for setting up the real contest – creating and
configuring accounts, defining languages, etc.
The only real difference is typically with the specification of the problem
set: the practice problem must be configured into the system for the
practice contest (it is undesirable to configure the real problems ahead of
time, as this would mean the problem names would be visible to the teams during
the practice). However, most
configuration items other than the problem set are usually exactly the same
during a practice contest as they are during the subsequent real contest.
At the end of such practice contest, it is necessary to “reset” the state
of the system by removing from the database all runs, clarification requests,
judgements, etc. which were submitted
during the practice contest. However, it
is at the same time desirable to avoid removing from the system the
“configuration information” such as account names, passwords, language
definitions, etc.
The Time/Reset
screen contains a function to support the operations required to move from a
practice contest to a real contest.
Pressing the Reset
Site button will clear the PC2 database at that site of
all runs, clarifications, and judgements from a practice contest. The Reset Site operation thus provides the ability to
clear all data from the practice session and leave the system ready for entry
of the real contest problem descriptions followed by a clean start of the real
contest.
Reset Site affects only the current site (the
site at which the Admin client is logged in).
Because the operation it performs has such a substantial effect, the
Contest Administrator is given a warning about the effects of the operation and
is asked to confirm the intent. The Administrator should only proceed if it is
clear that all activities related to the practice contest at the site are
completed. Also, the Reset Site button can only be activated when
the contest is in the “stopped” state at that site.
If the Contest Administrator confirms the Reset Site operation,
the system creates a backup zip file containing the practice contest state,
saves the zip file under the “archive” directory, and then clears the site
database of all information specific to the practice contest (runs,
clarification requests, etc.). This
leaves the system in a state where it is ready to have the problem descriptions
for the real contest entered, and then the real contest can be started.
The Reset All Sites
button has a corresponding effect on all connected sites (that
is, it notifies all other PC2 site servers that they should also
perform a Reset Site
operation). If the contest is still
running at another site when it receives such a “reset site” notification, the
contest is automatically stopped at that site.
The Contest Administrator is responsible for insuring that all
practice contest operations at all sites are finished prior to invoking
the Reset All Sites operation. The system provides one warning about the
significant effect of this operation, and then if the operation is confirmed it
instructs each site to perform a Reset Site operation (first stopping the contest at that site
if it is still running).
Note that both the Reset
Site and Reset
All Sites buttons at a given site are disabled unless the
contest is stopped at that site. However, as described above, when a
Reset All Sites operation is selected, it
is not a requirement that the contest already be stopped at other
sites. A Reset All Sites operation will stop the
contest at other sites automatically.
Use this function with caution!
Note also that the Reset
Site button is not the same thing as the “pc2reset” command
(see Built-in Commands, earlier in this manual). Whereas the Reset Site button clears data related to a prior
contest session, the pc2reset command completely
resets PC2 to an initial installation configuration – it is the same
as reinstalling the system from scratch. Specifically, for example, the pc2reset command deletes all account information, problem and
language descriptions, etc. Be sure to
use the proper command when attempting to “reset” your contest!
The Startup Status tab on the main Administrator screen is
used to track the status of Teams once a contest has been started. This is particularly useful during a
“practice contest” held just prior to a real contest; it allows the Contest
Administrator to verify that all teams have been able to login and use the
basic PC2 functions successfully.
A
sample Startup Status
screen is shown below. Initially all
teams are displayed in RED, indicating that the team has not made any contact
with the PC2 server. When a
team logs in, their display changes to YELLOW; when they have submitted at
least one run or clarification request their display changes to MAGENTA or BLUE
respectively; once a team has successfully submitted both a run and a
clarification the display changes to GREEN, indicating the team has
successfully performed all the basic PC2 functions and should be
ready to use the system in the real contest.
(The screen below shows teams in each of these states, although it may
be hard to read if you are not looking at a color copy of this manual.)
There is one
situation which can arise in a contest whereby the Startup
Status display will not accurately reflect
the contest state. If a team logs in
while there is no Admin client running, but does not submit any runs or
clarifications, then logs out before the Admin client logs in, then that team
login will not appear on the Startup Status display (the team
will show as “No Contact”). A team must
either already be logged in when the Admin client is started, or must login
while the Admin is logged in, or must leave at least one run or clarification
request in the system during the team login, in order for the Admin client to
recognize that the team has logged in.
Note that under normal circumstances an Admin client is always started,
and remains running, for the duration of a contest, so this situation will
rarely occur.
The Runs tab on the main
Administrator screen displays a grid showing all the runs which have been
submitted so far in the contest (from all teams, at all sites). The run display grid can be sorted on any
column by clicking in the column header; clicking multiple times toggles the
sort between ascending and descending order.
The columns can also be resized by moving the column separators in the
header. An example run grid is shown below.
The Runs display provides a number of capabilities for
the Contest Administrator. One function
of this display is to provide the ability to select a run which has already
been judged (and hence no longer appears on the Judge’s grid of available runs)
and “give it back to the Judges” so that it may be re-judged. (Note that this assumes an Administrative
decision to re-judge a run has been made for some reason; this is not a normal
contest operation.)
To give
a run back to the Judges, click on the row containing the run to select it,
then click the Give
button. This will cause the run to
appear on the Judge’s screens so that it can be selected and re-judged.
(Note: when a Judge selects a run which
has been sent for re-judging, a warning message is displayed on the Judge’s
screen asking for verification that the run really is intended for
re-judging.) A run does not disappear
from the Administrator’s grid when it is sent for re-judging; the Administrator
always has a complete listing of every run submitted in the contest, from all
teams at all sites.
A second purpose of the Runs display is to allow the Administrator to “take
a run away” from a Judge. This can be used, for example, to take back a run
which was given in error to the Judges for re-judging. Any time there is a run on the Judge’s
display grid which should not be there (because it has already been judged and
is not intended to be re-judged, for example), click on the run in the
Administrator’s Runs display then click the Take button. This will remove the run from the Judge’s
screens.
Another function of the Runs display is to allow the
Contest Administrator to edit various parameters associated with a specific run
– for example, to mark a specific run as “deleted” or to change the effective
submission time of a run. This allows
the Contest Administrator to make decisions regarding unanticipated situations
affecting how a run should be considered in scoring.
To edit a run,
select the run in the Runs
grid and then press the Edit button. This will bring up the following Edit Run dialog :
The “Problem”, “Language”, and
“Judgment” drop-down lists allow the Administrator to alter the specification
of the corresponding attributes associated with the run. It is allowable to change multiple attributes
of a run during a single edit (although this would be usual; normally, editing
a run is done for a specific purpose such as correcting a judging error).
Changing the “Problem” attribute will have the effect of changing the way
in which this run is considered in scoring:
it will be counted as a run for the newly specified Problem (however, this will only have an actual
effect on the scoring results if the Team has correctly solved the newly
specified Problem; see the chapter on the Scoreboard for details). Changing the “Language” attribute will have
the effect, if the run is subsequently re-executed, of changing the language
definition (compiler invocation) used to compile and then execute the run. Changing the “Judgment” attribute will have
the effect of causing the newly specified judgment to be the one used by the
Scoreboard in determining whether the Team has correctly solved this problem.
“Elapsed Time” represents the elapsed time in minutes from the start of
the contest (contest elapsed time, not
counting minutes during which the contest clock was stopped) at which the run
was received. The Administrator can
change this value as desired. Note that
Elapsed Time is considered the “team submission time” and is used to determine
the calculation of penalty points assigned to the run.
“Proxy Site” is an internal value used to indicate whether, in a
multi-site contest, this run has been handed off from its originating site to
another site for handling; this should normally not be changed. (See the chapter on Loosely-Coupled
Multi-Site Contests for further information on Proxy Sites.)
Checking the “Mark Run as Deleted” box will cause the run to be
completely ignored in all scoring computations.
Elapsed Time, Judgment value, and all other attributes of a run which is marked “Deleted” have no effect on scoring. (“Deleted” runs do not actually get removed
from the database (nor from the Runs
display), they are simply marked as such to indicate they should be
ignored for scoring purposes.) If a run
has been previously marked as deleted and subsequently the Mark Run as Deleted
box is unchecked, the run is once again considered in further scoring
computations, with no indication that it was previously “marked as deleted”.
The “Suppress Team Notification” checkbox is used to indicate whether or
not the Team which submitted this run should be sent a notification of the
“Judgment” value applied to this run after it has been edited. Normally, editing involves correcting an
internal administrative error and it is not desirable to notify the Team of any
editing, so the default operation is to suppress Team notification. If it is desired that the Team should
receive a notification of the result of editing the run (for example, a
judgment was changed from NO to YES and the Contest Administrator wishes the
Team to know this), uncheck the “Suppress Team Notification” checkbox.
The Execute
button on the Edit Run
dialog allows the Contest Administrator to execute a run just as it would be
executed on a Judge’s machine. This is
useful, for example, when a Team submitted a run with an incorrect language
specification (which most likely caused a Judge to render a “Compilation Error”
judgment for the run), and it is desired
to determine what the result would have been if the language specification had
been correct. The Language drop-down
list can be used to change the language attribute of the run, and it can then
be executed at the Admin workstation.
Note, however, that the Execute function will only work correctly on
the Admin workstation if the workstation has been configured with the necessary
language compilers, in the same way as on a Judge’s machine.
Once a run has been edited (and re-executed if desired), pressing the Update button will store
the new specification of the run’s attributes in the database and, if team notification
has not been suppressed it will
send a notice of the run status to the Team.
Note that once an update has been applied, the former state of the run
is lost; there is no way to restore a run to a prior state once it has been
edited (other than simply re-editing the run and changing the values back – but
PC2 does not keep track of the old run state once the Update button has been
pushed).
It is
frequently desirable to view a subset of the complete set of runs which
are currently in the database. For
example, the Contest Administrator may be interested in looking at all the runs
for a given Problem, or all the runs from a given Team, or all the runs
submitted during a specific window of time, or some combination of these.
The
Administrator Runs
display has associated with it a filter which can be used to apply a set
of filtering criteria to the runs which are displayed. Pushing the Filter button on the Runs display screen will produce the
following dialog, which is used to set the filtering criteria:
Selecting a set of items in the filter dialog indicates that those items should
be displayed on the Runs
grid. For example, to specify that the Runs grid should display
all (and only) runs from Team 1 at site “CSUS” for the problem named
“Penguins”, you would select “Penguins” in the Problems column, “Team1” in the
Teams column, “CSUS” in the Sites column, and “All” in the Language, OS, and
Time columns, and then press the Update button.
The Runs grid would then apply the specified
filter criteria to all runs, displaying only those that match the selected
criteria. As another example, to display
all (and only) runs which were submitted during the first 20 minutes of the
contest, you would enter “0” in the “From” time field, “20” in the “To” time
field, and select “All” in the Problems, Teams, Sites, Language, and OS fields,
and then press the Update
button.
The filter is designed with the expectation that at least one entry will
be selected in each column. The “Clear
All” button in each column is provided as a convenience to allow “deselection”
of all entries in anticipation of then selecting one or more entries in the
column. However, clearing all criteria in a column does not make sense as a
final filtering operation: if “Clear All” were allowed as the final choice
in a column (that is, if no entry in the column was selected following a “Clear
All” for the column), this would imply that no runs should be displayed
(since no runs would match the filtering criteria of “nothing selected”). For this reason, the filter ignores a “Clear
All” condition if it is the last operation performed on a column and instead
effectively selects all the criteria in that column. In other words, pushing “Clear All” as the
last operation on a column is effectively the same thing as pushing “Select
All” on that column.
The same logic applies to the “Clear All” button near the bottom of the
screen, which is used to initially clear all entries in all columns, in
anticipation of then selecting at least one entry in each column. If this global “Clear All” is the last
operation selected prior to pressing the Update button, the filter interprets this as a
request to disable the filtering operation entirely; that is, to display
all runs on the Runs
grid. Thus, the global “Clear All”
button is effectively a “Filter Off” (no filtering) button.
In order to remind the user when filtering of runs is taking place, any
time a filtering operation has been selected the “Filter” button on the Admin screen will change
color to blue and will indicate “on”, like so:
Filter (ON). Disabling filtering (using the “Clear All”
button as described above) will return the Filter button
to its normal state.
The Clars
tab on the main Administrator screen displays a grid showing all the
Clarification Requests which have been submitted so far in the contest (from
all teams, at all sites), in a format similar to the Runs grid. Like the Runs grid, the
Clarification Request grid can be sorted and resized by manipulating the
columns headers. The Give and Take function work like
the Give and Take on the Runs
pane. Note that there is no filter
operation supported for Clarification Requests.
The Reports
tab on the main Administrator screen provides a variety of options for
generating statistical reports about a contest, both during and after the
contest. The Reports screen
looks like the following:
The “Report Option”
drop-down list allows choosing one of a number of different report
formats. The list of available reports
is summarized in the table shown below.
Pressing the “View”
button will pop up a display showing the content of the selected report, and
will also write the selected report in text form to a file (located by default
in the $PC2HOME directory) using the file name given in the “Report File Name” text
box.
|
Option
|
Description of Report
|
|
-all
|
Details of all problems, languages, teams, and runs
|
|
-complete
|
Same as -all but dumps all reports as well
|
|
-runs
|
Details of every submitted run
|
|
-clars
|
Details of all submitted clarifications
|
|
-prob
|
Details of problem definition info
|
|
-1
|
Summary of submissions by language
|
|
-3
|
Number of teams that solved each problem
|
|
-4
|
Number of correct and incorrect solutions by problem
|
|
-6
|
Summary submissions, solutions, times by team/problem
|
|
-5
|
Summary of who should have balloons (“Balloon Report”)
|
|
-7
|
Run listing per team by problem
|
|
-8
|
Run summary grid
|
PC2 contains a separate “Scoreboard” module which keeps track
of the current standings in a contest.
The scoreboard provides several functions: it automatically generates HTML pages
describing the current state of the contest in a variety of formats; it generates email and/or printed “balloon
notifications” (when that option has been selected by the Contest
Administrator; see the section on “Options”, above); and it provides the capability to generate an
“export” file containing contest standings data (in the form required for
importing to the ICPC Registration system).
The HTML files generated by the scoreboard are placed in a directory
named “html” (lower case) directly beneath the $PC2HOME directory. Balloon notifications are sent to
destinations as configured by the Contest Administrator. The export data file is placed in the
$PC2HOME directory.
The following sections describe the details of the generated HTML files
and the export data file. They also
describe the overall scoreboard operation, including the scoring algorithm used
to compute contest standings.
To start a scoreboard, go to a
command prompt, be sure that the working directory is $PC2HOME (the PC2
installation directory) and type the command
“pc2board”. This will start a PC2 client
expecting a scoreboard login.
Once the Client login window appears, enter the scoreboard account name
“board1” and the board1 account password as defined when PC2
accounts were created,. This will bring up the PC2
Scoreboard display window (next page), indicating that the scoreboard program
is running.
The scoreboard automatically generates a complete set of HTML files to
the html/
directory as soon as it is started.
Thereafter it generates updated HTML files periodically according to an
algorithm described below. Each time a
new set of HTML files is generated the scoreboard display window is updated to
show the most recent update time.
Under normal circumstances it is only necessary to have a single
PC2 scoreboard running, even in a multi-site contest. The scoreboard automatically receives update
information from every site server, and
generates HTML files describing the
overall contest status (including all sites).
These HTML files can be copied to a publicly-accessible location for
access by a browser (see below), so participants at any location can see the
current standings. In addition, a
single scoreboard (running under the “board1” account) can generate balloon
notifications for all sites. Thus
there is rarely a need for running more than one PC2 scoreboard in a
contest, and this is the recommended mode of operation.
Once
it is running, the scoreboard waits passively until some contest event occurs
which could alter the data it should display (this could be a submitted run, a
judgement decision by the Judges, a change in a team name, or various other
events). When any such event occurs, the
scoreboard obtains from the server an update of the contest state, computes the
new standings based on this information, and regenerates the HTML display
files.
Because of the dynamic nature of scoreboard
updates and the large number of events which can affect the scoreboard status,
the scoreboard can generate a significant amount of communication traffic with
the server. Further, in a multi-site
contest, this traffic occurs with every
site server before the scoreboard can regenerate the display files. In a slow or crowded network environment, or
one subject to long delays, this communication overhead can be significant.
In
order to reduce the amount of scoreboard traffic, the scoreboard contains a
“throttling” mechanism to give the Contest Administrator control over the
frequency of updates. This mechanism is
based on a setting which can be specified in the [client' section of the pc2v8.ini file on
the machine running the scoreboard module. If the [client' section defines an attribute of
the form minSecsBetweenUpdates=N , then the
integer value N is
used as the minimum number of seconds which must elapse between scoreboard
updates. If there is no definition of minSecsBetweenUpdates in
the [client'
section of the pc2v8.ini
file, the default value for N is 180 (three
minutes).
The
value of N is used to control the scoreboard in
the following way. Every time an event
occurs which could potentially alter the scoreboard state, the scoreboard
checks to see if it has been at least N seconds since the last time it updated
the display. If N seconds have
not elapsed since the last update, the event is ignored for the moment. If N seconds HAVE elapsed, then the
scoreboard collects from all servers the current state of the contest and
creates a new set of HTML files based on that data.
Thus,
the default condition in PC2 is that the scoreboard is updated at
most every three minutes, but the Contest Administrator can arrange that this
time is increased or decreased as desired.
Note that setting the value of minSecsBetweenUpdates to
zero will have the effect of eliminating throttling; the scoreboard will compute new standings
every time any relevant event occurs (but at the expense of generating a
potentially significant amount of network traffic).
There
are two exceptions to the rule of limiting scoreboard update frequency. First, any time a “YES” judgement is issued
by the Judges, the scoreboard unconditionally updates, regardless of the
throttling value. Second, any event
which occurs while the contest is in a STOPPED state also will force an
unconditional scoreboard update (this includes the condition that the contest
is stopped because it has ended). In
all other circumstances the scoreboard updates in a time-constrained manner as
described above.
Each HTML file generated by the PC2
scoreboard is a complete stand-alone HTML document (i.e. is bracketed by <html>
… </html> tags). Each document <head> includes a <title> tag, into
which PC2 places the Contest Title as specified by the Contest
Administrator (see Options,
above). Each document <body> contains an imbedded <table> holding
contest status information.
The
<table> in
each different HTML file contains a different set of contest information, such
as team rankings, run submission statistics, etc. (see below). The information outside the <table>
can be edited/replaced as desired, for example by adding additional header
information, frames, or any other HTML constructs. However, it is important to keep in mind that
the set of HTML files is completely regenerated on every scoreboard
update; changes made manually to an HTML output file will only persist until
the next scoreboard update.
The following HTML files are always generated by the scoreboard:
|
File Name
|
Table Contents
|
|
full.html
|
Columns showing rank, team display name, number of problems
solved, and penalty points, with rows
ordered by rank. This is the standard
“contest standings” display.
|
|
fullnums.html
|
Same as full.html except
that the Team Display Name is preceded by the Team Number followed by a dash.
|
|
final.html
|
Standings as they are displayed
at the conclusion of the ACM ICPC World Finals – Penalty Points are
considered and displayed only for the top ten teams; the remaining teams who
have solved at least the median number of problems are grouped alphabetically
by number solved; the rest of the teams are listed as “Honorable Mention”.
|
|
sumtime.html
|
A grid showing, for each team
and each problem, the number of runs submitted by the team for the problem,
and, if the team has solved the problem, giving the contest elapsed time of
the team’s solution.
|
|
sumatt.html
|
A table similar to the sumtime table but instead of giving the time of
solution for solved problems simply indicates “Y” or “N” according to whether
the team has solved the problem.
|
|
summary.html
|
A table combining the full and sumtime displays described above – that is, a
showing rank, team display name, number of problems solved, penalty points,
and number of runs submitted and solution time for each problem, with rows
ordered by rank.
|
The most common way to take advantage of the HTML files generated by the
PC2 scoreboard is to run a separate external process (e.g. a batch
script) which repeatedly copies the current HTML files to some external
location, reformats them if desired, and makes them available to teams and
spectators using a browser.
There
are several advantages to this method of operation. First, the details of the appearance of the
scoreboard can be customized by the Contest Administrator external to PC2. Thus the Contest Administrator can choose to
take advantage of the full set of scoreboard screens, or can choose to omit
some or all of them. In addition, it is
not necessary for teams, judges, spectators, etc. to run separate PC2
scoreboards, since most users will already have access to a browser. The Contest Administrator can arrange that
the external scoreboard script builds the desired scoreboard display and puts
the resulting HTML in a standard public location accessible to all user’s
browsers.
In addition to the above files, the scoreboard can generate files showing
rankings based on the concept of “groups” or “regions” with which a team is
associated. For example, in the ICPC
World Finals, teams compete not only for placement in the overall world-wide
standings, but also among teams from their own region of the world for the
regional championship. Other examples include situations where it is
desirable to break contest teams up into separate groups based on level of
experience (e.g. “lower division” and “upper division” students), and in
multi-site contests where it is desirable to be able to display rankings that
show only those teams participating at a given site.
Every PC2 account has associated with it a “Region ID”
identifying the “region” or “group” to which the account belongs. By default
all accounts have a Region ID of zero (which is considered a valid Region ID); hence
all teams are by default in the same “region” (group). Changing the Region ID for a particular team
associates that team with other all teams having the same Region ID (see the
section on Editing Accounts, earlier in this manual).
By default, the scoreboard ignores Region IDs. However, it is possible to make the
scoreboard pay attention to Region IDs (that is, pay attention to team
groupings), and when this is done it will cause the scoreboard to automatically
generate additional HTML files, as follows:
|
File Name
|
Table Contents
|
|
regions.html
|
A single table showing the top
team in each region (team grouping).
|
|
regionX.html
|
This represents several
different HTML files, one for each region “X” (region1.html, region2.html,…. ). Each regionX.html file contains
the same contents as the file full.html described above, except that the data is
restricted to teams which are defined as belonging to “region X”.
|
There are two ways to cause the
scoreboard to pay attention to regional groupings of teams (and hence generate
the additional region HTML files). One
is to use the “import ICPC data” function on the main Administrator
screen. The imported ICPC data contains
information identifying for PC2 both the number of regions into
which teams have been partitioned, and the “region ID” and “region name”
associated with each team. This
information both enables the automatic generation of the regional HTML files,
and tells the system how determine the region with which a given team is
associated.
The
second method which can be used to enable regional HTML file generation is to
use attribute
assignments in the [board'
section of the scoreboard’s pc2v8.ini
file. Attribute values can be defined
for specifying the number regions, the name of each region, and the ID number
associated with each defined region.
If
generation of region/grouping information is enabled via pc2v8.ini attributes
(as opposed to being enabled via an Import ICPC data operation), the Contest
Administrator must also take care to assign a Region ID to each account
according to the region to which the account belongs. All accounts in the contest (even accounts
across multiple sites) which belong to the same region or group should have the
same Region ID number, since this is how the scoreboard determines the
association between an account and its region. (This assignment is handled
automatically when the Import ICPC data operation is used.)
For
example, if it was desired to separate all contestants into “Lower Division”
and “Upper Division” groups, all accounts used by teams in the “Lower Division”
might be given Region ID values of “1” while all accounts used by teams in the
“Upper Division” might be given Region ID values of “2”. Note that in a multi-site contest, all Lower
Division teams would need to have the same Region ID (2), regardless of the
site at which they were located.
As
another example, if it was desired to be able to display standings in a
multi-site contest on a per-site basis, the accounts for all teams at (for
example) Site 1 should be given Region ID values of “1” while the accounts for
all teams at Site 2 should be given Region ID values of “2”, etc. See the appendix on pc2v8.ini attributes
for further information.
The
algorithm used in PC2 to compute Rank and “Score” (Penalty Points)
is the one used in the ACM ICPC World Finals, which is as follows:
1)
Teams are ranked according to the number of problems
solved; a team solving more problems is always ranked higher than a team
solving fewer problems.
2)
Within a group of teams solving the same number of
problems, teams are ranked by increasing “Penalty Points” (that is, the team
with the lowest number of Penalty Points is ranked highest within the
group). Teams only accrue Penalty Points
for problems which the team has solved; unsolved problems do not affect
the scoring in any way. Teams accrue
Penalty Points for solved problems in two ways:
·
One point for each minute elapsed from the start
of the contest until the problem was solved (the time of SUBMISSION is counted
as the “time solved”; it does not matter how long it took the Judges to judge
it).
·
A specific number of penalty points for each
INCORRECT submission submitted to the Judges prior to a correct solution for the problem (runs submitted after a
correct solution are not counted in the scoring).
3)
If two or more teams have the same number of solved
problems and exactly the same number of
Penalty Points, ties are broken in favor of the team with the earliest
time of the last correct submission (that being the time when the team
“finished” the contest).
The
scoreboard provides the ability to generate a text file containing data
describing the current contest standings.
Since this capability was provided for purposes of exporting contest
results back to the ICPC Registration system, the button on the scoreboard
display which is used to invoke it is labeled “Export ICPC”.
However, the use of the export operation is not restricted to
interfacing with the ICPC Registration system; it simply generates a file
containing text records made up of comma-separated fields.
The
export data file contains, for each team, a record giving the number of
problems solved, the total number of penalty points accrued on solved problems,
and the time of last submission of a correct solution (used in the ICPC World
Finals as a tiebreaker). See the
Appendix on Import/Export Interfaces for
further details on the format of the export data file.
To
generate the export data file, simply press the Export ICPC button on the scoreboard display.
This causes the scoreboard to generate a file named “pc2export.dat” containing export
data. The file is located in the
$PC2HOME directory.
PC2 will support multi-site contests running in either
“tightly-coupled” or “loosely-coupled” mode.
Tightly-coupled mode means that there is an active network connection
between sites, whereas loosely-coupled mode means that there are separate sites
running the same contest but with no direct network connectivity between the
sites. In either mode, each site is
always running its own PC2 Server.
In tightly-coupled mode, all PC2 events which occur at any
site are automatically transmitted by the system to all connected sites. For example, runs submitted by a team at one
site automatically appear on the screens of judges at all sites; changes in the
scoreboard state at one site are automatically displayed at other sites;
etc. A tightly-coupled contest is
established by setting up one primary server together with one or more
secondary servers configured with pc2v8.ini entries which cause the secondary servers
to connect with the primary server at startup, as described earlier in this
manual.
If a single contest
is being run at multiple sites which do not have direct connectivity with each
other, it is possible to manage the overall contest by using PC2 running
in loosely-coupled mode at each site.
In loosely-coupled mode the servers do not directly communicate with
each other. Instead, each separate
site is carefully configured so that it is aware of the other sites in
the contest but does not directly communicate with them. To replace the functionality which is lost
due to servers not communicating directly with each other, PC2 provides
an off-line external mechanism for merging data from each site into a single
set of overall contest standings. This
means that every site can see the overall contest standings even though there
is no direct network connectivity between the sites.
In order to set up a loosely-coupled multi-site contest, the server at
each site must be made aware of the other sites, without being told to attempt
a network connection to the other sites. This is done by including in each server’s sitelist.ini file
a listing of ALL servers (sites) in the contest, while at the same time NOT
including a remoteServer=xxx
entry in any server’s pc2v8.ini
file. Thus every server is
operated as a “primary” server in a loosely-coupled contest.
Once a contest is set up to operate this way it is possible to arrange
for “off-line” communication and sharing of standings data between the sites
– including the ability to generate a
“merged scoreboard” showing the combined results of competition taking place at
multiple loosely-coupled contest sites. The following sections describe the
details and constraints involved in doing this.
To support loosely-coupled multi-site operation, it is useful to have a
separate site in the contest acting as the “Master Site”. The Master Site is responsible for off-line
acquisition of contest data from each of the other sites, for merging the data
into a “single contest” view, and for generating the “combined standings
scoreboard”. PC2 has built-in
functions to perform these operations automatically.
While it is theoretically possible to have one of the active competition
sites in the contest also act as the Master Site, we recommend instead that a
completely separate PC2
“site” be set up strictly for purposes of acting as the Master
Site. From the point of view of other
sites, this Master Site is just another site in the contest; it should be
listed in the sitelist.ini
file of all sites, just as all other sites must be listed. The Master Site can
be physically located at one of the actual contest sites; however it is
generally less confusing – and less error prone – if the Master Site is
logically distinct from an active competition site.
PC2 must be installed
and configured at the Master Site just like it is at all the sites where teams
are competing. The procedure for
installation and configuration of PC2 at the Master Site is
identical to that of any other site.
Once PC2 is installed,
start a PC2 server at the Master Site, then start an Admin
client. Login using the Admin root account
and configure the contest at the Master Site in the normal way. The Master Site must have the same PROBLEM
configuration as the other contest sites,
but it will not normally have any languages defined (since those may differ
between sites), nor will it normally have any Team or Judge accounts (it should
have Scoreboard and Admin accounts).
When a contest is run in tightly-coupled
mode, an Administrator at one site can effectively “start” the contest at all
sites simultaneously – by pressing the “Start All Sites” button on the
Administrator “Time” screen. This causes
PC2 to automatically start the “contest clock” on all servers, thus
synchronizing all sites. However, in loosely-coupled mode, there is no way
for the clock to be started simultaneously in PC2 since the servers
by definition cannot communicate with each other in loosely-coupled mode. This can create a very bad situation if
the humans involved do not take care to avoid it.
Specifically, if the contest data from various loosely-coupled sites is
to be transported to other sites using the Export/Import and Merge functions
(for example, in order to produce an “overall scoreboard”; see below), it is critical that all sites start their PC2
contest clock at the time the
contest starts at that site – meaning, some human at each site must push
the “Start Contest” button at the moment the contest problems are handed out
and the contest begins at that site.
Otherwise, even if all teams
receive the contest problems at precisely the same moment in real time and have
precisely the same amount of time to work on them, the merged scoreboard
results will be inaccurate due to incorrectly synchronized starts at various
sites!
For example, if all sites hand out the problems at the same instant but
one site doesn’t start the PC2
clock for another 30 minutes, then every run submitted at that site will
have a timestamp which is off (by 30 minutes) from the other sites. Since scores are computed based on times (and
hence timestamps), the merged scores will be wrong. We have seen this
happen – more than once – at loosely-coupled contests, and there is no way to
fix it except to edit all the run timestamps manually through the Administrator
interface at the Master Site.
Therefore we offer the following WARNING: The Contest Administrator needs to make sure there is someone at each loosely-coupled site that is
responsible for pushing the PC2 “Start Contest Time” button at the
moment the contest starts at that site.
Please see the additional discussion of this topic under “Starting the
Contest”, above. You’ve been warned….
The most common reason for using PC2 in loosely-coupled mode
is to provide the ability to generate a single scoreboard showing the merged
standings of all sites in the contest, even though the site servers do not have
active connections to each other. In
order to generate such a merged scoreboard, each site must “export” its current
configuration and standings to the “Master Site”, which “imports” the data from
each site and generates a single scoreboard from the merged data. This export/import/merge operation can be
done on the fly, at any time (and multiple times) during a contest.
To do the export, a Contest Administrator at each site must select the “Remote” tab on the main
Administrator screen, which produces the screen shown below. The Contest Administrator should insure the “Contest Configuration”
and “Standings”
items under “Items to Export” are both selected (checked), and then
click the Export
button. This will pop up a dialog
prompting for a file name into which the export data should be written. (Note: do not confuse this Export file,
generated for communication between loosely-coupled sites in a multi-site contest,
with the “ICPC Export Data” file described in the chapter on the Scoreboard and
in the Appendices; they are not related to each other.)
The “Select Export Target” list
on the right side of the Remote screen is not used when generating an export file
for merging standings.
Once the export data file is created it must be transferred by external
means (ftp, email, etc.) to the Master Site.
The mechanism by which this is accomplished is up to the Contest
Administrator; it is outside the scope of PC2.
Once the Master Site has received an exported file from a site, the file
is imported into the Master Site by clicking the Import button and selecting the file which
was received from the remote site. The
data in the file will be merged with the current standings in the Master
Site.
NOTE: it is important that each
site in a loosely-coupled multi-site contest be properly (identically)
configured in terms of the number and order of the contest problems. When an Import is done at the Master Site, the import
includes an update to the contest configuration as defined by the imported
data; if the data between various sites is inconsistent then the merged results
will be similarly inconsistent.
To generate a merged scoreboard (that is, a single scoreboard reflecting
the combined standings of teams competing at all “loosely-coupled” sites),
start a PC2 Scoreboard on the Master Site and login to a Scoreboard
account. This will generate standard PC2 HTML display files as described in the
chapter on the PC2 Scoreboard.
The data in the scoreboard HTML files will reflect the combined
standings of all sites whose data has been imported into the Master Site. The HTML files can then be published via a
web server, from which they can be accessed by teams at all sites, or they can
be transferred back to each individual site by external means (ftp, email,
etc.) and displayed locally by the site.
Many loosely-coupled
contests operate by having judges available at every site to handle submitted
runs and clarification requests from teams at that site. However, this is not always the case. Suppose that in a loosely-coupled contest
there are not judges available at all sites, or that for purposes of
consistency all judging is to be done by contest staff located at a single one
of the sites. Under these conditions,
teams at a site where there are no judges will submit their programs (to their
PC2 site server), but there will be no judge connected to the server
to accept the runs.
In addition to supporting merged-scoreboard
operations as described in the previous section, the Remote Export/Import
mechanism can be used to support cross-site judging in such situations. The Contest Administrator at the team’s site
can export the submitted runs from that site to a site where there are
judges. (Typically this would be the
“Master Site”, but it need not be).
Exporting team runs is done by selecting “Unjudged Runs” as one of the “Items To
Export” prior to pressing the “Export” button. Similarly, Clarification Requests which have
been submitted by teams (but have not been answered because there are no judges
at that site) can be exported to a site where there are judges, by selecting “Unanswered Clarifications”
prior to generating the export file.
In order for PC2
to be able to keep track of runs and clarifications which have been “exported”
for handling at another site, it is necessary that the export of runs and
clarifications be made “destination site-specific”. That is, when exporting runs and/or clarifications
it is necessary to identify precisely the site to which the runs/clarifications
are being exported. This is done by
using the “Select Export
Target (Server)” drop-down list to choose the site to which the
export data is going to be sent.
The receiving (target)
site in this situation is known in PC2 terminology as the “proxy
site” for the run/clarification, and the exported runs are said to be “proxied”
to the specified target site. The
generated export data file will be rejected during an import attempt by any
site other than the one for which is was targeted. Also, it is not allowed to “proxy” (export) a
run or clarification to more than one site at the same time. While this “targeted export” mechanism forces
somewhat of a constraint on cross-site operations, in that the Contest
Administrator must decide on a policy governing where exported
runs/clarifications are to be sent, the export/import facility still provides a
substantial contest management facility in an environment where there is neither
active network connectivity nor judging capability at all sites.
Once a set of exported
runs/clarifications has been imported into the destination site, the runs and
clarifications will immediately be displayed on the screens of all judges
connected to that site and can be handled by the judge just as if they had been
generated locally (i.e. by a team
connected to that site). Once the judge
finishes with the run or clarification, it goes back to the local server just
like any other submission. At this point
the Contest Administrator at the proxy site would use the Export function,
selecting “Completed Runs”
and/or “Answered Clars”
for export, and specifying the originating site as the target.
Once the exported file is
shipped back to the originating site and imported, the results (run judgements
and/or clarification responses) will be processed at the originating site just
as if the sites had been actively connected. In this way the Export/Import facility
provides a sort of “virtual connection” between otherwise disconnected sites in
a loosely-coupled contest.
The ability to
export/import runs can also be used to support a “verification” operation by a
“Head Judge” at the Master Site. Under
this scenario, the Contest Administrator would export from each site (and
import into the Master Site) the completed runs for each site. The Head Judge can then re-execute selected
runs as desired to verify correctness of the final standings.
Note however that in order to perform such a
“verification” operation, the Head Judge would need to be running on a machine
with a configuration which is identical to that of the team which submitted the
run at the remote site. If different
sites have different environments – some Windows and some Unix, for example – the
Head Judge will need to have access to both platforms in order to execute runs
in both environments.
Appendix A
– pc2v8.ini Attributes
As
described in the chapter on PC2
Initialization Files, the pc2v8.ini file consists of up to six sections: [server', [client'
, [admin', [judge', [team', and [board',
with each section containing one or more "attribute
assignment" statements of the form attributeName=value. Lines in the file which begin with a “#” or
“;” character are ignored, as are blank lines.
Attribute names (left side of the equal-sign) are not case sensitive;
however, string data on the right side of the equal-sign is case
sensitive.
The following list gives the
attributes which can be defined in each section of the pc2v8.ini file, along
with a description of their function.
Some attributes may appear in more than one section.
[server'
section attributes:
debugLevel=<integer>
Sets
the level of debugging output sent to the log file (pc2serv.log). Higher integer values cause more detailed
output to be written to the log. The
default value if this attribute is not specified is 2. Values above 99 are not useful (produce no
additional output). Values in the range
10-20 produce useful information without generating reams of internal data; a
good value for most contests is 13. If
you are having problems, raising this attribute value may help identify the
problem. This attribute can also appear
in the [client'
section.
consoleLevel=<integer>
Sets the level of logging output sent to the
server console screen. Higher integer
values cause more detailed output to be written to the screen. The default value if this attribute is not
specified is 2. Values above 99 are not
useful (produce no additional output). Values
in the range 5-20 produce useful information without generating reams of
internal data; a good value for most contests is 5. If you are having problems,
raising this attribute value may help identify the problem. This attribute can also appear in the [client' section.
site=<site_name_string>
Tells
the server the name of the site which it is serving. The value <site_name_string>
should be replaced with the exact character string identifying the Site (i.e.,
one of the lines from the sitelist.ini file). Site
names must be unique across all contest sites.
Note that site name strings are case-sensitive.
This attribute can also appear in the [client' section.
port=<portNumber>
Tells the server the port number on which it should
expect to be contacted by clients, and by other PC2 servers in a
multi-site contest. This attribute may
be omitted from the pc2v8.ini file, in which case it defaults to 50002. Note that if you choose to assign a specific
port number, then all clients and other servers contacting this server must
also be told to use this same port number (this is specified with the “server=” attribute in the case of
clients, and with the “remoteServer=” attribute in
the case of other servers). Note also
that if you choose to assign a specific port number, the port number should be
greater than 49151 according to the conventions established by the IANA
(Internet Assigned Numbers Authority).
remoteServer=<IPAddress>:<portNumber>
Tells the server the IP address and port number of a
remote PC2 server at another site which it should contact in order
to join a multi-site contest. The <portNumber> must be
specified and must match the port number being used by the server at the remote
site. The appearance of this attribute
makes this server a “secondary” server; if this attribute is not defined in the
[server' section then
this server is a “primary” server and waits passively to be contacted by other
site servers.
[client'
section attributes:
site=<site_name_string>
Tells
the client the name of the site to which it belongs. The value <site_name_string>
should be replaced with the exact character string identifying the Site (i.e.,
one of the lines from the sitelist.ini file on the server). Site names must be unique across all contest
sites. Note that site name strings are case-sensitive. This attribute can also
appear in the [server'
section.
server=<IPAddress>:<portNumber>
Specifies the IP address and
port number at which the client module should attempt to contact the PC2
server. Every client module MUST have a “server=<IPAddress>:<portNumber>” entry in the [client' section of its pc2v8.ini file. The IP address and port
number must correspond to the address of the machine running the PC2 server and the port number at which the server on that machine is
expecting to be contacted.
workDir=<fullPathName>
Specifies the “working directory” for the
client. This directory is used as the
default directory in dialogs which require the user to select a file. If this attribute is not specified, the
default is the directory from which the client was started. This attribute can also appear in the [team' section.
minSecsBetweenUpdates=<integer>
Specifies the number of
seconds which the scoreboard should wait before performing and update when an
event occurs. Events which occur in less
than the specified amount of time since the previous event are ignored, except
for certain key events such as the occurrence of a “YES” judgement which always
force a scoreboard update. The
default value for this attribute if it is not specified is 180 (three
minutes). Setting this attribute to have
the value zero effectively forces the scoreboard to update on the occurrence of
any event. (See the chapter on
the scoreboard for information on the ramifications of setting this attribute.)
minSecsBetweenUpdates=<integer>
Specifies the number of
seconds which the scoreboard should wait before performing and update when an
event occurs. Events which occur in less
than the specified amount of time since the previous event are ignored, except
for certain key events such as the occurrence of a “YES” judgement which always
force a scoreboard update. The
default value for this attribute if it is not specified is 180 (three
minutes). Setting this attribute to have
the value zero effectively forces the scoreboard to update on the occurrence of
any event. (See the chapter on
the scoreboard for information on the ramifications of setting this attribute.)
[board'
section attributes:
regionIdCount=<integer>
Specifies the number of distinct
groupings or “regions” into which contest teams have been assigned. If this value is greater than zero it causes
the scoreboard to generate “region” scoreboard HTML display files for each
defined “region” (group) of teams. The
default value for this attribute is zero,
which suppresses the generation of region files by the scoreboard. If this attribute is defined then the regionNameCount attribute (below) must also
be defined. See the section on the
scoreboard for further details.
regionNameCount=<integer>
Specifies the number of
distinct region names which are defined.
This attribute must be defined if regionIdCount (above) is defined, and it must have the same value as regionIdCount.
regionIDX=<integer>
Specifies the ID number
associated with region “X”. Note that this represents the occurrence of multiple attribute assignments, one for
each defined region. In other words,
there must be exactly as many regionIDX attribute assignment statements in the pc2v8.ini file as the value of the <integer> which appears in the regionIdCount statement (see above). The integer which appears on the right side
of a regionIDX statement can be any
non-negative value; different regions are not required to have adjacent integer
values (only unique ones). Note,
however, that the values of “X” on the left side in the
set of regionIDX attributes should start at 1 and have
increasing consecutive integer values (see the example below).
regionNameX=<string>
Specifies the name
associated with region “X”. Note that this paragraph represents multiple attribute assignments, one for
each defined region. In other words,
there must be exactly as many regionNameX attribute assignment statements in the pc2v8.ini file as the value of the <integer> which appears in the regionNameCount
statement (see above). Different regions
should have unique name strings.
To clarify the use of region
groupings, the following example is provided.
This set of pc2v8.ini entries will cause the scoreboard to generate regional
HTML files for two regions: one consisting of all teams whose Region ID is
three (grouped into a region named “Upper Division”), the other consisting of
all teams whose Region ID is seven (grouped into a region named “Lower
Division”) :
regionIDCount=2
regionNameCount=2
regionID1=3
regionName1=Upper Division
regionID2=7
regionName2=Lower Division
[admin'
section attributes:
useTeamName=<anything>
useSchoolName=<anything>
useSchoolShortName=<anything>
useTeamAndSchoolName=<anything>
These four attributes are
used in conjunction with importing ICPC Registration data into PC2. If any one of these attributes is defined
(that is, if the attribute name on the left side has been assigned some
non-empty string on the right side), then the Admin client uses the
corresponding item from imported ICPC data to determine the Display Name which
will be used for a team on the scoreboard.
For example, if “useSchoolName=true” was specified (and none of the other attributes were specified), then the team’s full school name would be
used on displays. The default value if
none of these attributes is specified is useTeamName. If more than one of these
attributes is specified then the highest one in the list as given above has
precedence. These attributes are only
useful if ICPC import data has been used to set the various choices of names
available for display. See the section
on importing ICPC information for further details.
[judge'
section attributes:
defaultAnswer=<string>
Specifies the string which will be sent to a Team
(from a Judge) in response to a Clarification Request if the Judge pushes the
“Default Answer” button rather than typing in a specific answer to the
Clarification Request. The default value
for this attribute if it is not defined (and hence also the default value for
the Judge’s “Default Answer” button) is “No Response”. Note that this attribute must be defined in
the pc2v8.ini file for every judge who will be
responding to clarifications.
[team'
section attributes:
workDir=<fullPathName>
Specifies the “working directory” for the team. This directory is used as the default
directory in dialogs which require the team to select a file. If this attribute is not specified, the
default is the directory from which the Team client was started. This attribute can also appear in the [client' section.
Appendix
B –
Networking Constraints
As mentioned in the beginning of
this manual, PC2 modules must be able to communicate with each other
via TCP/IP: clients must be able to
communicate with their servers, and servers in a multi-site contest must be
able to communicate with other servers.
If client machines reside on the same LAN segment as their server, and
if all servers have publicly routable IP addresses which can be reached by
other servers, then communication should work with no problems.
However, given the wide variety
of network configurations which can exist – firewalls, NAT, and VPNs, just to
name a few – there may be some constraints in a given network setup which will
cause problems in setting up a contest using
PC2. In order to
understand how to avoid (or circumvent) these problems, it is useful to have
some understanding of how PC2 networking is implemented.
PC2 is written in Java and uses
Java RMI (Remote Method Invocation) for communication between modules. RMI is a high-level remote-communication
protocol which encapsulates information regarding connections between separate
network nodes (machines) and allows a Java program running in a Java Virtual
Machine (JVM) on one node to make direct calls to methods in a program running
in a separate JVM, typically on a separate network node.
When a PC2 server is
started, it “registers” itself with a process running on the same machine,
known as the Java RMI Registry (or just “registry” for short). If there is no registry running on the
machine when the server starts, then the server arranges to start the registry.
Registering the server with the
registry includes notifying the registry of the port number at which remote PC2
modules (clients or other servers) expect to make initial contact. This initial contact port number is obtained
by the server from the pc2v8.ini file. Registration also includes obtaining back
from the registry a port number at which the server should listen for remote
connections. This is the port number
that will be given to any remote modules which contact the registry asking for
a connection to the server. This port
number is generated automatically (internally) by the Java RMI Registry; PC2 has no control over the port
number.
Subsequently, when a remote PC2
module (a client or another server) wishes to contact the PC2
server, it first contacts the RMI Registry on the server’s machine. The registry in turn gives the client (or
other server) a handle to the initial server.
This handle contains the IP address and port number for the server; the
remote module can then directly communicate with the server using the IP
address and port number provided by the RMI Registry. Note again that this mechanism is part of the
Java RMI protocol; it is outside the scope of PC2 and PC2
has no control over it.
There are several important
ramifications of the RMI protocol implementation in terms of effects on PC2
networking. First, it means that PC2
will not work if there is a firewall between the server and the remote module
attempting to contact it and the firewall is blocking access to the port(s)
which the RMI Registry has assigned for communication between the server and
the remote module.
Note that this does not
refer to the port number which the Contest Administrator specifies in the “port=”, “server=IP:port”, and “remoteServer=IP:port” entries in the pc2v8.ini file. Those
port numbers are effectively the port at which the server notifies the RMI Registry
to listen for connection attempts. The
RMI Registry essentially “hands off” any such connection attempts to a (random)
port, notifying both the server and the remote module of that new port number. Since PC2 has no control over the
generation of that new port number, it means that the firewall must allow
connections between any port number which could be generated by the RMI
Registry.
Another constraint on networking
has to do with NAT (Network Address Translation). If one PC2 module is operating from
behind a NAT box, then when the RMI Registry issues a handle to a remote module
giving the IP address of the first module, that IP address will be the
“untranslated” address – i.e. the “local address” behind the NAT box. When the handle is eventually received at a
remote destination on the other side of the NAT box, the local address will not
have been translated by NAT (since the address is encapsulated inside an RMI
packet). This means that the remote
destination will have no idea of the correct “return address” (the public
address to which addresses behind the NAT box are translated) to use for
communication. In other words, it is a
limitation of RMI protocol that it does not work through NAT. Thus it is not possible to run PC2
modules when the communication needs to go through NAT to succeed.
While RMI protocol will not work
through NAT, it will work across a VPN (Virtual Private Network). This is a potential solution to both the
firewall and NAT problems. If the
Contest Administrator arranges for a VPN connection between PC2
modules, those modules will be able to communicate using RMI just as if they
were on a public network. An explanation
of how to set this up is beyond the scope of this manual, but we have tried it
and it works; a good network administrator should be able to set up such a
VPN.
Appendix
C –
Restarting A Server
As described in the section on Server Startup, when a PC2 server is started the
user will be prompted with a question, as follows:
Are there already servers up (Y, N, or Q) ?
If this is a case of restarting a server which was previously
running in a multi-site contest but crashed or was shut down for some
reason, and if there are other servers
still up and running in the contest, then this is said to be a case of a server
attempting to rejoin a contest.
Special steps are necessary in order to allow a server to rejoin a
contest.
First, since modules in the current version of PC2 do
not have the capability of detecting a loss of network connectivity, the servers which are still running will
think the server for the site being restarted is still “connected” (even though
it has been shut down). The running
servers will therefore refuse to accept a connection from the rejoining
server. Thus, before the server being
restarted can rejoin the contest, the other servers must be told to drop the
(bogus) connection to that server. This
operation is referred to as a “forced logout” of the disconnected server.
To perform a forced logout on the server being restarted, it is
necessary to use a PC2 “Admin
client” connected to some active server.
To do this, start a PC2 “Admin client” connected to an active
site, by setting the “site=” and “server=” entries in the [client'
section of the pc2v8.ini file for the
Admin client so that they point to the active server at another site. Next, go to the Sites tab on the main Administrator screen, and if the server
is listed as “active”, select the site and click on the Logoff button. It
is only necessary to perform the forced logoff operation on one active server;
that server will automatically notify the other servers to drop the dead server
connection.
Once the active servers have dropped the dead connection to the
server being restarted, that server can rejoin the contest. To do this, first answer “Y” to the question
shown above. At this point one of two conditions holds for the restarting
server:
- The
server’s pc2v8.ini file indicates it is a secondary server, by virtue of containing an
active “remoteServer=” entry. In this case the server will automatically attempt to
contact the primary server at the IP:port address given in the remoteServer entry in the pc2v8.ini file.
Assuming the primary server is up, the restarting server will
automatically rejoin the contest.
- The
server’s pc2v8.ini file indicates that it is a primary server, because it does not contain
any remoteServer entry. In this case the
server will issue a second prompt:
Enter
Server to Contact:
In
this case the user should enter the IP address and port number, in the form “IPAddr:port”
(without the quotes), of some
other server which is currently up and running in the contest. The reason this is necessary is that otherwise
the restarting server, since it is ‘primary’ according to its pc2v8.ini file, would simply sit and wait to be
contacted by other servers – which would require shutting down and restarting
the (secondary) servers still running at other site(s).
Note: an alternative way to accomplish the same
thing (that is, to allow a primary server to rejoin a running contest) would be
to edit the server’s pc2v8.ini file prior to the restart and add a new remoteServer=IP:port entry, thus effectively switching this
server from primary to secondary status.
This would effectively cause the entire contest to be running nothing
but “secondary” servers. This would not
be a problem as long as no complete restart is needed; as mentioned previously,
once the system is up and running in “steady state”, all servers act as peer
equals. If a complete restart became
necessary, it would require at least one of the site servers to be
re-designated as “primary” so that it could be started and wait for other
(secondary) servers to contact it.
Rejoining a contest can cause confusion if the Contest
Administrator who is restarting the server forgets to insure a “forced logout”
of the dead connection. In this case,
the active remote server will refuse to accept the connection from the
restarted server. This in turn causes
the restarting server to log a message describing what has happened, and then
simply give up (terminate). If you are
attempting to restart a server and rejoin a contest and it seems to simply quit
for no reason, check the contents of the file pc2serv.log. If the log ends with a series of messages
like
Error Contacting/logging-in to other server
Server from [this site' already logged in??
Server halted, have someone logoff this site.
then the server to which you are trying to
connect thinks you are already connected and you need to perform a “forced logout”
of your server connection on the running server(s).
Appendix
D –
PC2 Server Command Line
The PC2
server is a Java class named “Server” in a subpackage named “server” of the
package “pc2”. The command to start a
server is
java pc2.server.Server
typed in the $PC2HOME directory. This is in essence the command invoked by the
PC2 “built-in” command pc2server.
The server accepts a
number of command line options when it is started. One option is -h (or --help), which produces the following “usage” output:
$ java pc2.server.Server
--help
CSUS Programming Contest
Control System
mailto:pc2@ecs.csus.edu
Version 8.4 20030501 01
(May 1st, 2003 3:12pm) Java ver 1.3.0_02
Usage: Server
[-v|--version|-h|-help' [-first|-join' [-site sitename'
-h - this message (--help)
-v - print version information (--version)
-first - no prompt, this
server is a first server (no join)
-join - no prompt, this server contacts a remote
server
-site sitename - add to contest with site named sitename
As seen from the
“usage” output, the server also accepts the following command line options:
-v (or
--version) : prints the version
identification for the current PC2 server
-first : indicates that this
server is a primary server and should not attempt to contact any other servers
(ignores any remoteServer=
attribute in pc2v8.ini)
-join : indicates that this
server is a secondary server and should attempt to contact a remote server to
join a contest. Uses the remoteServer=
attribute in pc2v8.ini,
if there is one, to determine what remote server to contact; if no attribute is
defined, prompts at the console for IP:port connection information
-site <sitename> : sets
the sitename for the site being served by this server (overrides any
specification of site name in pc2v8.ini )
Appendix E
– ICPC Import/Export Interfaces
E.1 Importing
ICPC Registration Data
As mentioned earlier in this manual,
PC2 was designed for supporting the ACM International
Collegiate Programming Contest, including its local and Regional contests
worldwide. The ICPC maintains an online
Contest Registration system which is used by Regional Contest Directors (RCDs)
around the world to manage participation in the various ICPC Regional Contests. PC2 provides interfaces to import
contest registration data from the ICPC Registration system, and also to export
contest results back to the ICPC web site.
To import ICPC Registration data to PC2, the RCD must first
log into the ICPC Registration system and download the “PC2
Initialization” zip file which is automatically created and updated as
registration data changes occur. Once the PC2 Initialization data
file is downloaded, it should be “unzipped” at any convenient location.
Unzipping the PC2 Initialization data file will produce three
separate files: “PC2_Contest.tab”,
containing details about the organization of the contest (such as the formal
name of the contest); “PC2_Site.tab”,
containing data identifying the sites in the contest; and “PC2_Team.tab”,
containing data about the teams that are registered in the contest.
The PC2_Team.tab file
consists of text-based tab-delimited records, one record for each team
registered in the contest. The
tab-delimited field contents of each record are as follows:
1) An integer
giving the ICPC Team ID (note that this is not the same thing as the PC2
Team ID; see below).
2) An integer
giving the ICPC Region ID (the “region” or “group” to which the team belongs).
3) A single
character indicating the Team’s “registration status” for the contest – typically
either ‘P’ (pending), ‘A’ (accepted), or ‘C’ (cancelled). PC2 ignores records containing ‘C’
in this field.
4) A string
giving the Team Name; for example, “The
Top Coders”.
5) A string
giving the full name of the Team’s school; for example, “California State
University, Sacramento”.
6) A string
giving the Team’s school name in “short form”; for example, “CSUS”.
7) A string
giving the Team’s school’s URL; for example, http://www.csus.edu.
8) A string
giving the Team’s school’s country code (three letters); for example, “USA”.
9) A single
character ‘Y’ or ‘N’ indicating whether the Team’s school has a graduate
program . PC2 ignores this
field (but it must be present).
PC2 can “import” this team data and use it to define things
such as the team’s Display Name on the scoreboard. PC2 will associate consecutive
records in the PC2_Team.tab
file with consecutive PC2 accounts – the first record in the file
will automatically be associated with “team1” at the site specified in the
record, etc. (Note that PC2
does not use the ICPC Team ID field for purposes of identifying a team.)
If the contest administrator wishes to associate registered teams with PC2
accounts in a different order, the PC2_Team.tab file can be edited (rearranged) to
present the records in the desired order – the team which should be “team1” in
PC2 should appear first in the file, etc. Alternatively, if the PC2_Team.tab file
is copied to a new file named “_PC2_Team.tab”,
the new file can be edited by adding a new leftmost column containing a PC2
team number. PC2 checks for
the existence of this file and, if it is present, uses that file not only to
load the ICPC data but to determine, based on the value in the first field in
each record, to which PC2 account each record in the data applies.
Once the PC2 Initialization data has been unzipped and the PC2_Team.tab file
edited if desired, the next step is to press the “Import ICPC Data” button on
the PC2 Administrator main screen (under “Manage Accounts” on the Accounts tab pane).
This will cause PC2 to
display a “file selection” dialog; this dialog is used to identify for PC2
the set of .tab files which should be imported.
To make this identification, simply navigate to the directory containing
the unzipped ICPC PC2 Initialization data files, and select any one
of the “.tab” files in that directory.
This will cause PC2 to load all three of the initialization
files. Note: all initialization “.tab”
files must all reside in the same directory.
If this is the first time an ICPC
Import has been done, PC2 will show a dialog displaying the names of
the contest sites as defined in the ICPC data and asking the administrator to
provide a specification of the association of those site names with the site
names which PC2 already knows about (from its sitelist.ini
file). Once this is complete, PC2
saves the region-association information
and then finishes the association of ICPC Registration data with PC2
accounts; the results can be viewed on the Accounts tab on the Administrator main screen,
just as if the data had been entered manually.
(Note: the account display on the Administrator screen is not
dynamically updated upon completion of an ICPC import; click to another view
and then back to the “Manage Accounts” view in order to see the imported data.)
Note that each record in the ICPC PC2 Initialization data
contains multiple names associated with each team – the team name, the
full name of the school, and a short version of the school name. PC2 can be configured to use any
one of these names, or a combination of them, as the name to be displayed on
the scoreboard. See the appendix on pc2v8.ini file entries, under the [admin' section, for details.
PC2 expects quotation marks in the team data to be
“quoted”. That is, if any field in the
data contains a quotation mark (˛),
then (1) the entire field must be surrounded by an additional matching set of
quotation marks, and also (2) each quotation mark which is part of the data
must be doubled. Thus for example a team
name like The ˛TOPS˛
Team should appear in the
import data file as ˛The ˛˛TOPS˛˛ Team˛. If the PC2
Initialization file exported from the ICPC web site contains data with quotes,
it may be necessary edit the data by hand (or load it into a program such as
Microsoft’s Excel and then save it) to insure that quotation marks are properly
formed. If this is not done prior to importing the data into PC2,
you may see “format error” messages in the log file, and the data containing
the quotes will not be displayed properly.
E.2 Exporting
Contest Data
Pressing the “Export ICPC” button on
the scoreboard display causes the scoreboard to generate a text file containing
the current contest standings in a form required by the ICPC World Finals
results display algorithm. This file is made up of a series of text
records, one record for each team in the contest. Each record contains a set of comma-separated
fields. The fields in each record are:
1)
ICPC Team ID (note: not the PC2 team
number).
2)
Rank in contest (this field is blank if the team falls
in the “Honorable Mention” category as defined by the scoring display algorithm
for the ICPC World Finals; see the description of the file final.html in the chapter
on the PC2 Scoreboard for details).
3)
A non-negative integer giving the number of problems
the team has solved.
4)
A real number giving the total number of penalty points
accrued by the team.
5)
A real number giving the time of the last submission by
the team, taking into account only the problems which the team has solved (used
as the ICPC World Finals tiebreaker determination).
The records in the file are sorted by team rank based on problems solved
and penalty points. In particular, teams
that fall into the “Honorable Mention” category as defined by the ICPC World
Finals results display algorithm will still appear in rank order in the file,
as defined by number of problems solved, penalty points accrued, and
tie-breaking time of last submission.
The
exported data file contains all the information necessary to update the ICPC
web site registration system with the results of a contest. It is primarily intended to provide an
automated mechanism for Regional Contest Directors to post the results of
Regional Contests. The export data file can also be imported
into any program that wishes to make use of the standings data.
Note that if no “ICPC Import” operation was performed prior
to invoking the “Export
ICPC” operation, then PC2 will have no record of the ICPC
Team ID associated with each team. In
this case the “Team ID” value in field #1 in the exported file will be zero.
This problem can be circumvented in contests other than Regional Contests (that
is, contests where there is no ICPC data to import) by creating a local version
of the “PC2_Team.tab”
file, entering the appropriate PC2 team account number
in lieu of the ICPC Team ID. This will
cause the “Export ICPC”
operation to generate a file containing all the data necessary to compute
complete contest standings utilizing PC2 account numbers.
Appendix F –
Validators
F.1 Overview
PC2 allows the Contest Administrator to configure each problem
so that it has associated with it a validator program whose purpose is
to help automate the judging process. A
“validator” is a program which is given, as input, the output of the execution
of a run (that is, the output of a program submitted by a team).
Validators can operate in one of two ways: passively or actively. A passive validator is a program which
accepts the team program’s output and displays it in some useful form for
examination by the human judge. An
active validator is a program which not only accepts the team program’s output
but contains logic designed to make a determination, according to some set of rules,
regarding the correctness of the output.
An active validator can also return the result of its determination to
PC2.
An example of a passive validator would be the use of a “side-by-side”
file comparison utility which displays the team program output beside an
“answer file” provided by the Judges ahead of time. This allows the judge to use the validator to
examine the output and compare it with the correct answer, perhaps taking
advantage of special capabilities of the validator program (for example,
syntax-driven color highlighting of differences between the team program’s
output and the correct answer file).
When the judge is satisfied, the validator is terminated and the judge
uses the PC2 Judge client interface to enter a judgement.
An active validator is one which automatically examines the team program
output and makes some determination of its correctness. The validator must
contain program logic which directs how it determines correctness. This logic could be hard-coded within the
validator (in which case the validator is almost always problem-specific), or
could be more general (for example, instructions directing application of
“difference testing” between the program output and a Judge’s “answer
file”). In either case an active
validator is one which is designed to make an “automated” determination of the
correctness of the output of a team’s program.
For active validators, a set of conventions defined within PC2
provide a mechanism for the validator to return to PC2 an indication
of what judgement it thinks should be applied to the run. However, such a validator response is never
blindly accepted as the final determination of the judgement to be applied to a
run. Rather, the validator output is displayed
to the human judge as a “recommendation”.
It is ultimately up to the human judge to accept the validator
recommendation (or not). That is, it
still requires the judge, at a minimum, to press a button to accept the
validator recommendation.
F.2 Validator Configuration
By default there is no validator attached to (associated with) a contest
problem in PC2. Validators
can be attached to a problem by the Contest Administrator by using the Validator tab on the Edit Problem dialog. This displays the Validator configuration
screen, as shown below. When the Contest Administrator configures a problem to
use a validator, then when a Judge executes a team program the specified
validator will automatically be invoked as soon as the team program completes
execution.
The Validator screen allows the Contest Administrator to choose one
of two options for attaching a validator
to the problem being defined: either the
simple built-in PC2 validator, or an arbitrary separate external
program. If a separate external program
is selected it can be either a standard system program (such as a “diff”
utility or a split-screen file-compare utility) acting typically as a passive
validator, or it can be a program
written specifically for the validation of output from programs for this
particular contest problem, i.e. an active validator.
The Contest Administrator can choose whether or not to display the
validator result to the judge. Checking
the box “Show Validation To Judges
(SVTJ)?” when configuring a validator will cause the response returned by the
validator to appear on the Judge’s display when the run finishes
executing. If the checkbox is
unchecked, the validator result will not
be visible to the judge. The checkbox should
normally be unchecked when using passive validators, which do not return
meaningful result information (if SVTJ is checked and a passive validator is
used, the response displayed to the judge will always be “Undetermined”).
The following sections describe the built-in PC2 validator, the use
of an existing external program as a validator, and the creation and use of a
special-purpose problem-specific validator.
For problem-specific validators, the ICPC standard for interfacing such
validators to a contest control system such as PC2 is described.
F.3 PC2
Built-In Validator
PC2 contains a built-in validator which can be
used to compare the output of a submitted program with the “answer file”
provided by the judges (specified when a contest problem is configured). The built-in validator is essentially a very
simple version of the standard Unix “diff” file comparison utility. Selecting the “Use PC^2 Validator”
option enables a drop-down list showing the various comparison options, which
are listed in the following table.
|
PC2
Validator Option
|
Meaning
|
|
1
|
Perform a standard
“diff”: a character by character comparison of files differences
|
|
2
|
Ignore whitespace
at start of file
|
|
3
|
Ignore leading
whitespace on lines
|
|
4
|
Ignore all
whitespace on lines
|
|
5
|
Ignore empty lines
|
|
6
|
2 then 3
|
|
7
|
2 then 4 (implies
3)
|
|
8
|
2 then 5
|
|
9
|
3 then 5
|
|
10
|
4 then 5 (implies 2
then 3)
|
When using the PC2 built-in validator it is
important to understand the exact meaning of the options which specify
combinations (options 6 and
higher). In particular, “X then Y” means
that the validator first checks the output using case X and then checks
the output using case Y, independently.
Note that this is not the same thing as performing a single test
applying both conditions simultaneously.
This can produce unexpected results.
For example, “check the
output, ignoring white space at start of file” followed by “check the output,
ignoring empty lines” is not the same as
“check the output ignoring white space at start of file and also ignoring empty
lines”; if there are empty lines in the output file but not in the answer file,
then when it validates using the criteria “ignore white space at start of file”
the output will FAIL due to the empty lines in the answer file. Be sure you understand the operation of the
built-in validator before using it!
The PC2 built-in validator is an
active validator – meaning that it automatically returns to the Judge a
recommendation for the judgement which should be applied to the run whose
output was diff’d against the answer file.
If SVTJ was checked when the validator was attached to the problem, then
the returned recommendation is displayed for the judge.
F.4 Using An Existing External Tool As A
Validator
If the Contest Administrator selects Use External Validator on the Validator
configuration screen, it tells PC2
to use an external program as a validator for the problem being
configured. In this case PC2 needs to be given two sets of
information regarding the external program which is to be used: (1) the name of the file containing the
validator, and (2) the command sequence which PC2 should use when
invoking the validator.
To understand how this works, suppose it is desired to use an external
program named “gvim” as a validator for a certain contest problem. Further, suppose gvim is a program
which can be given an argument, “-d”,
along with two additional arguments specifying file names, and that what it
does in this case is display the contents of those two files side by side,
highlighting the differences.
What the Contest Administrator wishes to do is have PC2 invoke
gvim every time a run for this particular problem is submitted and
executed by a judge, and pass to gvim two files: (a) the output of the team’s program, and (b)
the answer file (correct output) for the problem (provided ahead of time by the
judges). This will allow a judge to compare
the team’s output with the correct output, using the capabilities of the gvim
program to assist them.
Since our hypothetical gvim program already exists and is not
(likely) written to understand how to make programming contest-related
decisions about the correctness (or not) of the differences between two files,
we will use gvim as a passive validator (one which does not return a
value to PC2). In that case
the Contest Administrator would uncheck the SVTJ box, and then would
select Use External Validator
on the Validator screen.
In the Validator Program textbox the Contest
Administrator enters the file containing the gvim program. This might be something like
/usr/local/bin/gvim
or
C:\Program
Files\gvim\gvim.exe
Use the “browse” button (the button showing “…”) to select the validator
program file name. Typically this is an
executable program, but it could be any file.
A legitimate file must be selected even if this field is not actually
used.
In the Validator
Command Line textbox the Contest Administrator enters the command to
be used by PC2 to invoke the validator. This command can include parameter
substitutions similar to those allowed when configuring languages. As when
defining languages, parameter substitutions are indicated by a set of matching
curly braces with a colon as the first character and containing a substitution
keyword, for example, {:infile} The following table shows the substitutions
which will be performed by PC2 prior to invoking the validator:
|
Keyword
|
Meaning
|
|
validator
|
Represents the file name given
in the Validator Program
box. Note that it is not a required
that the “validator program” actually be an executable program; it could for
example be a text file.
|
|
infile
|
Represents the problem data
input file as configured in the problem.
|
|
outfile
|
Represents the output sent to stdout
by the team program when it was executed by the judge.
|
|
ansfile
|
Represents the judge’s answer
file as configured in problem.
|
|
resfile
|
Specifies the name of the file
into which an active validator must place an XML representation of the
judgement.
|
Thus to invoke gvim as
described above the proper Validator
Command Line entry would be:
{:validator} -d {:outfile}
{:ansfile}
This would invoke the gvim program
( {:validator} ) with a “-d” argument, passing it
the team program output ( {:outfile} ) and the judge’s correct-answer file ( {:ansfile} ).
Note that while the above example uses an existing program (gvim)
as the “validator program”, it is not a requirement that this actually be an executable
program. As an example, the “validator
program” file might be a text file containing a set of “rules” describing how
to determine the correctness of a team’s output, and a completely separate
program could be invoked via the Validator Command Line entry, perhaps passing the “validator
program” file as a parameter.
For example, suppose a contest problem required teams to write a program
to generate output which conformed to a set of lexical rules. If there existed a program named “analyze”
which performed lexical analysis of the contents of a file according to a set
of rules specified in another file, then the Contest Administrator might create
an appropriate set of rules (the rules to which the team program’s output must
conform) in a file named rules.dat, select rules.dat as the “Validator Program” file,
and then specify the following as the Validator Command Line entry:
analyze {:validator} {:outfile}
This would invoke anaylze as the program to be executed during the
validation step, passing it rules.dat
and the team output file as input. (What
analyze would do with this is left as an exercise for the reader – but
see the following section.)
F.5 Implementing a Validator
The use of an existing system tool (e.g. gvim) as a validator has
the drawback that such validators typically must be passive, since they
normally have knowledge of neither how to make decisions regarding correctness
of team program output nor how to return such decisions to PC2 even if they did know how to
make them. The built-in PC2
validator can be used as an active validator since it knows how to make
decisions and knows how to return them, but as mentioned previously it is an
extremely weak implementation.
Thus, it is desirable to have a mechanism which allows the use of a
special-purpose program written specifically to be an active validator. In order to support the use of such
validators, it is necessary to define a set of standards for interfacing
validators to the system. Because a
validator is a problem-specific (not contest-specific or contest control-system
specific) entity, ideally these standards should be uniform and general enough
that a conforming validator (and its corresponding contest problem) could also
be used in conjunction with contest control systems other than PC2.
Under the auspices of the ICPC, the PC2 development team has
worked with a number of other contest system development teams to define such a
standard. Known as the ICPC
Validator Interface Standard, it specifies two interfaces: the
invocation interface from a contest control system to a validator program, and
the result interface from the validator back to the contest control system.
The following subsections give a brief summary of the interfaces defined
by the standard,
followed by a description of the PC2 implementation of the
standard. Since PC2 complies
with the standard, any validator written to comply with these interfaces can be
used as an active validator for a problem in PC2.
F.5.1 Invocation Interface
The
standard mandates that the contest control system is responsible for invoking
the validator and passing it at least four command line parameters, as follows:
parameter1: a string specifying the name of the input data file
which was used to test the program whose results are being validated.
parameter2: a string specifying the name of the output file which
was produced by the program being validated when it was run using the data file
named in parameter1 (that is, the name of the file containing the output to be
'validated').
parameter3: a string specifying the name of an arbitrary
"answer file" which acts as input to the validator program. The
answer file may, but is not necessarily required to, contain the "correct
answer" for the problem. For example, it might contain the output which
was produced by a "Judge's Solution" for the problem when run with
the data file named in parameter1 as input.
Alternatively, the "answer file" might contain information, in
arbitrary format, which instructs the validator in some way about how to
accomplish its task.
parameter4: a string which specifies the name of the "result
file" which the validator must produce. The content of the result file
produced by the validator is defined in the following section.
The
requirements for passing parameters to a validator can be met by the Contest
Administrator in PC2 through
the use respectively of the {:infile}, {:outfile}, {:ansfile},
and {:resfile} command substitution parameters in the Validator Command Line;
PC2 will automatically insert the appropriate values for the
problem when the validator is invoked.
Also, as required by the standard,
the data file, program output file, and the answer file are in the
current directory when the validator
program is run. These conditions taken
together specify how an arbitrary validator program should expect to be invoked
by PC2.
The standard
specifies that the contest control system may pass additional command line
parameters to a validator, as long as the first four command line parameters
are specified as listed above, and that the interpretation of any such
parameters is up to the validator. The
Contest Administrator in PC2 can pass arbitrary additional
parameters to the validator by including them in the Validator Command Line after the four
required parameters.
F.5.2 Result
Interface
The
standard requires that the validator result be returned in the “result file”
whose name is specified by parameter4 (above), and that the contents of the result file must
be a valid “XML Document”. This means
that it must start with a valid XML declaration,
such as
<?xml
version="1.0"?>
The root element of the XML
document must be of the form
<result outcome = "string1"> string2
</result>
The tag name “result” is fixed and required
by the standard, as is the attribute name “outcome”.
“string1” is an “outcome
string” defining the result (outcome) which the validator is reporting to the
contest control system. If the value of
“string1” is
“accepted” (or any case-variation of that word), the standard specifies that
the validator is indicating that the program output file “passed” the
validation test(s). If “string1” contains any
value other than a form of the word “accepted”, the standard specifies that the
validator is indicating that the program output file “failed” the validation
test(s).
In PC2, the appearance of any form of the word “accepted” in
the “string1”
attribute of the result element in the result file causes PC2 to
assign a recommendation of “YES” to the run being executed. Recommendations are displayed to the judge if
the SVTJ checkbox has been checked when the validator is configured with the
contest problem.
If
the value of “string1”
is not some form of the word “accepted”, then PC2 compares the
actual string value to the “reject messages” originally defined in the reject.ini file. If “string1” matches one of the reject messages, then
PC2 assigns that message as the recommendation for the run being
executed; otherwise, it assigns a recommendation of “Undetermined” to the run.
“string2”
is an arbitrary message string being returned from the validator to the contest
control system. The standard specifies
that the interpretation of this string is up to the contest control system. PC2 does not use the “string2” parameter from
the result file.
F.5.3 PC2 Extensions
The
standard specifies that the XML <result> element
produced by the validator may include other attributes in addition to the
“outcome” attribute, and may also include additional (nested) elements; it also
specifies that the interpretation of any such additional attributes and/or
elements is up to the contest control system.
Such additional attributes can be used to implement a variety of
features.
PC2 makes use of additional
attributes to implement a form of security.
Specifically, it expects the validator to define an additional attribute
named “security”
and to return in that attribute the name of the result file. That is,
PC2 expects the XML result file to look like:
<?xml version="1.0"?>
<result outcome="string1" security="resfile"> string2
</result>
where
“resfile” is the
value which was passed to the validator as the name of the file into which the
results should be placed (and where string1 and string2 are as described above).
Each
time PC2 invokes a validator it generates a unique random name for
the result file. When the validator
returns PC2 examines the contents of the result file and verifies
that the security
attribute value matches the file name.
Since a user (team) program cannot know ahead of time what result file
name PC2 will generate, it is not possible for a user program to
generate a “fake” result file which somehow gets used in place of one generated
by the validator. While this is not a
complete guarantee of security, it does make it much more difficult for a user
program to circumvent the operation of the validator.
Appendix G –
Language Definitions
As described earlier in this manual, PC2 must be given a
“language definition” for each language to be used in the contest (that is, for
each tool which teams can use to write and submit programs). The language definition consists of four
distinct text strings: the “Display
Name”, the “Compile Command Line”, the “Executable Filename” specification, and
the “Program Execution Command Line”.
In order to help in understanding how such language definitions work (and
so that you will be better able to develop your own language definitions), it
is useful to understand what it is that
PC2 does with a language definition. Language definitions are used by PC2 in two circumstances: when a Team invokes a TEST RUN operation, and
when a Judge or an Admin invokes an EXECUTE operation. The following algorithm describes the
sequence of steps which PC2 follows when either the TEST RUN button on the
Team, or the EXECUTE
button on the Judge or Admin is pressed.
1.
The entire contents of the “execute” directory (beneath the $PC2HOME
directory) are deleted. If something
prohibits this clearing, the system stops and displays a warning message, and
all remaining steps are skipped.
2.
The submitted files are copied to the execute directory.
3.
If the file whose name is specified as the “Executable
Filename” in the language definition exists in the execute directory, it is deleted. This prohibits a team from submitting an
executable file (or more correctly, they can submit it but it will never be
executed).
4.
The command specified as the “Compile Command” in the
language definition is executed, using appropriate command parameter
substitutions as defined earlier in this manual.
5.
PC2 checks for the existence in the execute directory of a
file whose name matches the specified “Executable Filename”. If this file exists, it must have been
created by the execution of the “Compile Command”. (This is how PC2 determines
whether compilation was successful.)
6.
If the specified “Executable Filename” exists (hence,
the “Compile Command” was successful), then the following operations are
performed:
a.
The data file associated with the problem (if any) is
copied into the execute
directory.
b.
The command specified as the “Program Execution Command
Line” is executed, using appropriate command parameter substitutions as defined
earlier in this manual.
c.
If this is an EXECUTE operation on a Judge or Admin (as
opposed to a TEST RUN on a Team), then if there was a “Validator” associated
with the problem, then the following
operations are performed:
i.
The “answer file” associated with the problem (if any)
is copied into the execute
directory.
ii.
The command specified as the “Validator Command Line”
is executed, using validator command parameter substitutions as defined earlier
in this manual (see the Appendix on Validators).
iii.
If “Show Validator Result To Judge” (SVTJ) was checked
when the Validator was associated with the problem, PC2 reads the
result file created by the Validator (see the Appendix on Validators) and
displays the appropriate result for the Judge.
Some examples of PC2 language definitions which have been used
in past contests and were known to work in those environments are shown
below. Each definition consists of four
lines, corresponding to the four text field entries required on the Edit Language screen when
defining a new language under the main Administrator screen.
No guarantee is made that these definitions will work in your
environment, nor that they will not become obsolete due to changes made by the
various language tool vendors. All we
can tell you is that all of these language definitions have been used
successfully in past contests. Use them
at your own risk.
Language: Java
Java
javac
{:mainfile}
{:basename}.class
java
{:basename}
Language: GNU C++
GNU
C++
g++
-lm -o {:basename} {:mainfile}
{:basename}
./{:basename}
Language: GNU C
GNU C
(gcc)
gcc
-lm -o {:basename} {:mainfile}
{:basename}
./{:basename}
Language: Borland Kylix C++
Kylix
C++
bc++
-A {:mainfile}
{:basename}
./{:basename}
Language: Borland Kylix Delphi (Pascal)
Kylix
Delphi (Pascal)
dcc
{:mainfile}
{:basename}
./{:basename}
Language: Microsoft Visual C++
MSVC++
cl
{:mainfile}
{:basename}.exe
{:basename}.exe
Language: Turbo Pascal for DOS
Turbo
Pascal
tpc
{:mainfile}
{:basename}.exe
{:basename}.exe
One of the ramifications of the sequence of language-handling steps
described above is that a team cannot submit a program whose file name is the
same as the “Executable Filename” specified in the corresponding language
definition. For example, if the Contest
Administrator configured a language by saying that the result of a compile
operation for the language was to produce an executable file whose name was
always “a.out”,
then if a team submitted a source code program in a file named “a.out”, then the source code program file would get
deleted (step 3) prior to the compile step.
Normally this difficulty is eliminated through the use of command parameter substitutions; the Contest
Administrator would not normally specify “a.out” as the expected executable file to
be generated by the compilation steps, but rather would use a specification
such as “{:basename}.out”,
and further a team would normally submit source code in a file named, e.g. “a.c” rather than
“a.out”.
However, there is one scenario under which the mechanics of language
handling by PC2 can cause difficulties (or at least,
confusion). This is the case of purely
interpreted languages, such as Perl or shell-script. In these cases there is no “compilation” step
which is expected to generate an “executable” file; the “source file” is
effectively the same as the “executable” file (in the sense that the source
file undergoes no transformation prior to invoking “execution”, since
“execution” involves running an interpreter against the original source
program).
For example, in the case of Perl, the Contest Administrator might attempt
to configure the language definition as:
Perl
/bin/perl -c {:mainfile}
{:mainfile}
/bin/perl {:mainfile}
This definition says that the language Display Name is “Perl”; that the
“Compile Command” invokes /bin/perl
( the Perl interpreter) with the “-c” (check syntax) argument and the submitted file,
that the result of “compilation” is to produce an “executable” file whose name
is the same as the submitted file, and that following the “compilation” step PC2
should check for the existence of the submitted file and if present it should
invoke /bin/perl again,
this time executing the Perl commands in the submitted file.
However, this language definition will not work, because of the steps
which PC2 follows: it will
delete the submitted source code (.pl) file prior to invoking the compilation command.
Again, the reason for this is that it checks for the existence of the specified
“executable file” after the compilation step, and assumes that if the
file exists then the compilation was successful. Thus if a team submitted a source file named
“myFile.pl”,
since the submitted file is the “executable file” which would be input to the
Perl interpreter, the “executable filename” (after command parameter
substitution) would also be “myFile.pl”
– but the file would have been deleted.
It is still possible to use such
languages with PC2. The trick
is to create a separate “script” file which acts as the “Compile Command” and
has the effect of creating a separate file which has the same name as that specified
for the “Executable Filename” and which PC2 can test for after
compilation and prior to invoking program execution.
For example, suppose the Contest Administrator creates a shell script
file named “compilePerl”
with the following contents (the example presumes a Unix-like environment, but
a similar approach can be taken in a Windows system):
#!/bin/csh
perl –c $*
if ($? == 0) then
touch OK
endif
This script basically says: run
the “C-Shell” interpreter (line 1); have it execute the perl interpreter and
perform a syntax check (-c)
on the arguments passed to the script ($*) (line 2); check the system “status
variable” ($?) and if it is zero (meaning no errors occurred) (line 3) then create a file named “OK” (line 4).
With this compilePerl
script accessible via the PATH variable, the following language
definition will allow a Perl program to be submitted and processed by PC2 :
Perl
compilePerl {:mainfile}
OK
perl
{:mainfile}
This language definition will invoke the compilePerl script telling it to
syntax-check the submitted program file, then if the file “OK” exists (which
will only happen if the Perl syntax-check was successful) it will invoke the
Perl interpreter to execute the submitted program. Note that while this example is for Perl,
other languages such as Bourne Shell (and other shells), Python, Ruby, and ‘awk’ can also use a similar solution.
The above example should provide some insight into the types of
operations which the Contest Administrator can invoke from PC2 . For example, it is possible to create a
script file which is invoked for the “Program Execution Command” and does any
desired operation, such as copying a data file into the execute directory
prior to running the intended program.
Basically any desired operation can be performed at either the “compile”
or “execute” step, as long as one has a clear understanding of the PC2 language
processing algorithm described above.
This organization of language processing gives a great deal of
flexibility to the Contest Administrator.
Appendix H – Extending PC2
While
the client interfaces (Admin, Judge, Team, and Board) in PC2 are
intended to be as general as possible, there may be situations where users
would like a client to operate differently.
For example, a user may wish to create a scoreboard that uses a
different scoring algorithm, or to create a different sort of contest system
interface for Teams. To support this, PC2
provides a mechanism for users to create their own “custom clients” which
interface with the rest of the PC2 system.
All clients in PC2 are extensions of a Java class named BaseClient. The BaseClient class defines an interface between the
PC2 server and an arbitrary graphical user interface (GUI). As a result, any class which extends BaseClient can be built
to provide any desired GUI, and can use the methods in BaseClient to obtain
access to the PC2 server.
Because BaseClient
is a general-purpose interface to the server, it contains many methods which
are not needed (or desirable to have) in certain kinds of clients. To further simplify the process of creating
custom clients, PC2 defines three abstract subclasses of BaseClient: TeamClient, JudgeClient, and ScoreboardClient. Each
of these abstract classes defines specifications (prototypes) for the methods
which a client of the corresponding type needs to implement. Thus for example a user interested in
creating a new type of Team client should extend the abstract TeamClient class (rather
than directly extending BaseClient),
and implement the abstract methods specified in TeamClient.
The BaseClient class, along
with all its direct and indirect subclasses, is part of a Java package named “ex” (“extensions”) which
in turn is a subpackage of package “pc2” (thus the full class name for BaseClient is pc2.ex.BaseClient). The downloadable PC2 distribution
contains a directory named “doc” in which exists a subdirectory named “pc2ex” containing the
“Javadoc” documentation for the entire “pc2.ex” package; opening
a browser on the file “index.html”
in the $PC2HOME/doc/pc2ex
directory will display the documentation.
Users who are interested in additional information on developing custom
clients should contact us through the PC2 web page for further
details.
Appendix I –
Frequently Asked Questions
This appendix covers some
of the common questions we get. For
additional information, visit the PC2 web page FAQ at http://www.ecs.csus.edu/pc2/doc/faq/index.html.
Question:
Why can't my PC2 clients connect to the PC2 server when I
run the server on my Linux machine? (It works fine when the server is running
under Windows, even if the clients are on Linux machines.)
Answer: We have seen this happen when a Linux
machine running the PC2 server has not been properly configured for
a network environment. Specifically, Linux machines typically start up
configured to report their "machine name" as "localhost".
PC2 queries the OS to obtain the machine name, and then maps this
name to a corresponding IP address. The machine name "localhost" maps
to the "loopback" IP address 127.0.0.1. This ends up being the IP
address which the PC2 server broadcasts to the clients as the
address they should use for communication back to the server -- which means the
clients will be unable to contact the server because they will use their own
local loopback channel.
To check whether
your Linux machine running the PC2 server has this problem, type the
command "hostname" (or "/bin/hostname") at a command
prompt. If this returns "localhost", your OS is not properly
configured for network operations.
One way to fix
this problem on most machines is to edit the file /etc/sysconfig/network and change the
HOSTNAME entry to a host name which resolves to a non-local IP address, and
then reboot the machine.
Another approach
is to use the "hostname" command, run as root, to set the hostname to
either a specific non-local IP address or to a hostname which resolves to a
non-local IP address. One user with this problem wound up changing the hostname
before starting PC2, via a command line like:
hostname $(ifconfig eth0 | grep inet | sed
's/.*r:\| B.*//g')
or (depending on
your shell)
hostname 'ifconfig eth0 | grep inet | sed
's/.*r:\| B.*//g''
which is
basically getting the address from the eth0 interface and setting the hostname
to that value. In any case, be sure to coordinate such actions with your local
network administrator.
Question:
What does it mean when the system generates the message
"NoClassDefFound" and won't run ?
Answer:
This is a Java error message meaning that it could not find one of the
"classes" (programs) required to run the system. The missing classes
could either be PC2 modules or Java system modules. There are a
number of possible causes for this error. Here are some things you can try:
Verify that the PC2 system was
unzipped "including subdirectories" and "with
case-sensitivity". PC2 uses many Java "packages",
which must be stored in the proper subdirectories beneath the PC2
installation directory. Make sure the program which was used to
"unzip" the distribution properly created and maintained the
hierarchical directory structure and the "case" of filenames in the
system.
Verify that the CLASSPATH environment
variable includes the Java/lib and PC2 installation directories.
