An error in the design of the dependent tuples functionality was discovered. See the mailing list archive and this Wiki page for details.

With the default set of modules the problem only manifests itself when certain combinations of the fixedtime.so and consecutive.so or sametimeas.so modules are used. In those cases Tablix 0.3.1 may produce timetables that do not respect some restrictions given in the problem description file. Earlier versions are not affected since they do not support updater functions. The current CVS version of Tablix does not correct this issue, but should give a warning about a failed assertion if the final timetable is incorrect (Tablix must be compiled with the --enable-debug option for this warning to be displayed).

The part of the kernel API that deals with dependent tuples and updater functions will most likely be redesigned in the next release to fix this problem.

Denis Voskvitsov discovered a bug in Tablix 0.3.1 which causes a division by zero if a resource domain in a problem description is empty. This condition can for example arise from a "capability" restriction for an event that does not match any "capability" restrictions for classrooms and causes Tablix to lock up a few moments after starting (no progress indicator or error messages are shown). Tablix 0.3.0 most probably also has a similar bug.

This problem is now fixed in the CVS version.

Jorell front end for Tablix is accepting your timetabling problem descriptions again.

The server-side script is now more robust and can better tolerate errors on the computing cluster back-end running at Cyberpipe. I also modified some error messages to give a better explanation of what exactly is wrong with the submitted XML file.

However please note that this is still an experimental service. Bug reports and suggestions are welcome as always.

Debian development packages once again contain a properly compiled CVS version of Tablix.

The problem was caused by a missing pvm-dev package on the machine that is doing the weekly automatic builds. Because the configure script could not find a properly installed PVM library it configured Tablix for debug mode and I didn't notice that in the build logs.

I have noticed that Tablix web front end at jorell.tablix.org began to behave strangely. It is reporting errors about unknown Tablix parameters even for perfectly valid problem descriptions. The cause of the problem currently seems to be at the computing cluster and not at the server front end. Nonetheless this service will now be off-line until I can locate the bug and fix it.

On a possibly related note it also came to my attention today that the automatic build system for Debian packages is broken. It appears that the development packages at http://www.kiberpipa.org/~tomaz/tablix/devel contain a build intended for debugging purposes that is not using PVM. The packages containing stable builds are not affected.

Sorry for the inconvenience.