Dear Hubert, Op 28 jan. 2014, om 09:38 heeft Hubert Garavel <hubert.garavel@inria.fr> het volgende geschreven:
Thanks for your answer dated May 2, 2013. I apologize for the delay in responding but after submitting the bug, I did not receive any answer in my mailbox. I now understand that the answer was posted to the MRMC mailing list, to which I did not register, so I only saw your anwser today by chance browsing the mailing-list.
Indeed, I agree that the .tra file was not a correct input file for MRMC, because it was in PRISM format, which is similar yet slightly different.
The point of my bug report was different: given a wrong input file, MRMC should emit an error message, instead of crashing with a memory problem *** mrmc: double free or corruption (!prev): 0x09e3b008 ***
This was really the essence of my bug report.
You are right. I sent another mail to Joost-Pieter with a proposed fix of the bug. It basically is:
In the branch https://svn-i2.informatik.rwth-aachen.de/repos/mrmc/branches/nijmegen-small-..., I added more strict tests of the file contents that produce an error message whenever a line does not follow the format exactly. This branch also includes a number of other improvements to the MRMC code, e.g. stricter type checking, a corrected Fox-Glynn algorithm (see my report https://pms.cs.ru.nl/iris-diglib/src/getContent.php?id=2011-Jansen-Understan...), and more efficient solving of linear equation systems.
It also includes my implementation of sparse matrices (see Appendix D of the MRMC user manual, also available as a patch from http://www.mrmc-tool.org/downloads.php), which we did not insert in the trunk because simulation became inefficient. Shortly after releasing MRMC 1.5, I found the error and corrected that inefficiency.
Sorry for the delay... I had thought that this fix would be quickly included into the trunk of the svn repository by someone in Joost-Pieters group. (If the person responsible for the repository desires, I can merge the mentioned branch into the trunk. The branch passes all tests. However, I will not produce a full new release.) Kind regards, David.