Version 2.1.38 (20050607)
=========================

libchecker
----------
The checker programs do not terminate with the "Check Failed"
error, if they are given more command line parameters, than are
required. The extra parameters are ignored. This enables use of
the standard checkers (for example, cmp_int) for checking
problems, which have tests with .inf files.

run,serve
---------
The interface between the `run' and the `serve' programs is
reworked again. The `run' program always generates the testing
protocol in the XML-based format. Generation of the
HTML-representation of the testing protocol is now performed by
the `serve' contest server. Because of this there is no need in
two separate protocols: master's and team's. Both may be
generated from the XML testing protocol. The XML testing
protocols are stored in the subdirectories of the
"${contest_root}/var/archive/xmlreport" directory. Backward
compatibility with previous versions of the ejudge is provided:
if for some submit there is no XML testing protocol, the plain
text or HTML testing protocols are used (generated by the
previous versions of the ejudge tools).

Because the `run' program always generate the testing protocol
in the XML format, the `html_report' global configuration
variable is no longer meaningful. This variable is preserved for
compatibility with previous versions, but its value is ignored.

Added possibility of storing the full program's output in the
contest archive. By default the input test data, the program's
output, the checker's output for each test are added directly to
the testing protocol. However, the maximal file size and the
maximal string length of such file are limited (see the
max_file_length and max_line_length global configuration
variables). If the actual file size is greater than the file
size limit, the file content is not added to the protocol, but a
short diagnostic message is added instead. This version of the
ejudge system supports `enable_full_archive' global
configuration variable. If this variable is set to 1, the output
of the program being tested is not limited. The output is stored
in a separate file rather than in the testing protocol. The
testing protocol does not contain neither input, nor output data
of the program being tested, with the exception of the command
line parameters, if such are provided. The program output and
the checker program output are stored in a separate file in a
special archive format, which is currently a custom format, not
compatible with any existing archiver. The archive contains some
file information and the file data compressed with the gzip
algorithm. The archive files are stored in subdirectories of the
"${contest_root}/var/archive/output" directory.

serve,master,judge,team
-----------------------
Since in the full archive mode the testing protocol does not
contain the input data and the output data, added the
possibility of viewing archive input and output files. The
output file, the standard error stream, the checker output for
each test are taken from the archive file. The input file, the
correct answer file (if provided for a problem), the info file
(if provided) are taken from the test directories for the
problem. It is possible, that the test directories contain
different tests from those used for actual testing (for example,
if the tests were updated from the time of testing). The current
version of the ejudge does not check correspondence between file
versions.

serve
-----
An audit log is kept for each submit. The audit log contains the
information about all operations with the current run. The audit
log is never overwritten or edited, new entries are only
appended to the end of the log. The audit log files are kept in
subdirectories of the "${contest_root}/var/archive/audit"
directory.

serve,master
------------
Viewing of the audit log is supported. In order for a user to
view audit log, the user must have the CONTROL_CONTEST bit set
for the current contest.

serve,master,judge,team
-----------------------
Support for the obsoleted session modes is removed. The
obsoleted session modes are login/password in the HTML form,
cookie, session id in the HTML form. These session modes are not
used for a long time, but were supported in the code. Their
removal considerably simplified the source code. The URLs are
also simplified, because the sid_mode CGI parameter is no longer
required.

userlist-server
---------------
The program does not crash on login of a user, which has no
password set (i.e. the `register_password' entry is completely
missing).