Before the contest starts, a number of things will need to be configured by the administrator. You can check that information, such as the problem set(s), test data and time limits, contest start- and end time, the time at which the scoreboard will be frozen and unfrozen, all from the links from the front page.
Note that multiple contests can be defined, with corresponding problem sets, for example a practice session and the real contest.
The problem sets are listed under `Problems'. It is possible to change whether teams can submit solutions for that problem (using the toggle switch `allow submit'). If disallowed, submissions for that problem will be rejected, but more importantly, teams will not see that problem on the scoreboard. Disallow judge will make DOMjudge accept submissions, but leave them queued; this is useful in case an unexpected problem shows up with one of the problems. Timelimit is the maximum number of seconds a submission for this problem is allowed to run before a `TIMELIMIT' response is given (to be multiplied possibly by a language factor). Note that a `timelimit overshoot' can be configured to let submissions run a bit longer. Although DOMjudge will use the actual limit to determine the verdict, this allows judges to see if a submission is close to the timelimit.
Problems can be imported and exported into and from DOMjudge using zip-files that contain the problem metadata and testdata files. See appendix DOMjuge problem format specification. Problems can have special compare and run scripts associated to them, to deal with problem statements that require non-standard evaluation. For more details see the administrator's manual.
The `Languages' overview is quite the same. It has a timefactor column; submissions in a language that has time factor 2 will be allowed to run twice the time that has been specified under Problems. This can be used to compensate for the execution speed of a language, e.g. Java.
DOMjudge comes with some tools to check for mistakes in the
testdata. These tools are all located in the misc-tools
directory in the source tree.
The 'checkinput' programs are meant to check testdata input (and optionally also output). They check for simple layout issues like leading and trailing whitespace, non-printable characters, etc. There's both a C program and AWK script which do essentially the same thing. See 'checkinput.c' for details. All scripts take a testdata file as argument. The 'fixinput.awk' script corrects some of these problems.
This program can be used as a more advanced replacement of checkinput.
It allows you to not only check on simple (spacing) layout errors, but
a simple grammar file must be specified for the testdata, according to
which the testdata is checked. This allows e.g. for bounds checking.
See appendix
checktestdata-grammar
for a grammar specification.
Two sample scripts checktestdata.{hello,fltcmp}
are provided
for the sample problems hello
and fltcmp
.
This program is built upon the separate library libchecktestdata.h
(see checktestdata.cc
as an example for how to use this library)
that can be used to write the syntax checking part of special
compare scripts: it can easily handle the tedious task of verifying
that a team's submission output is syntactically valid, leaving just
the task of semantic validation to another program. When you want to
support `presentation error' as a verdict, also in variable output
problems, the option whitespace-ok
can be useful. This
allows any non-empty sequence of whitespace (no newlines though) where
the SPACE
command is used, as well as leading and trailing
whitespace on lines (when using the NEWLINE
command). Please
note that with this option enabled, whitespace matching is greedy, so
the script code
INT(1,2) SPACE SPACE INT(1,2)
1__2
(where the two underscores
represent spaces), because the first SPACE
command already
matches both, so the second cannot match anything.
Before a contest, you will want to have tested your reference solutions on the system to see whether those are judged as expected and maybe use their runtimes to set timelimits for the problems. There is no special method to test such solutions; the easiest way is to submit these as a special team before the contest. This requires some special care and coordination with the contest administrator. See the administrator's manual for more details.
If your contest has a test session or practice contest, use it also as a general rehearsal of the jury system: judge test submissions as you would do during the real contest and answer incoming clarification requests.