Klever

BTW, because of this issue I got wall timeouts for verification tasks requiring ~7.5 minutes of CPU time with timelimit equal to 15 minutes. This depends on a server load and hardware characteristics, but this is very bad from the benchmarking point of view. We need to find any solution. Perhaps we can disable wall timeouts in BenchExec and enforce them in scheduler workers ourselves.

This issue is not strictly bound with purposes of 2.0 that targets verification of C software.

Implemented in 6614-speculative-scheduling.

There is little sense to proceed to testing and for merging this branch to master as it needs again a way to install and to configure non-standard versions of BenchExec. I won't be surprised if everything will fail with current automatic deployments.

It is implemented now.

• ef86f4f2d3ca1c1df0344dbc89d830f6,
• 0d990ceb2317b87546c35448b3635305,
• c450089f71843085027c5f38f9796a63,
• f0a889ca491447c356784de52b21520c,
• f4a68caa2c6b9c490e93591c21e0bbb1,
• 789b222612c061dc0d8286eb19de1fd1,
• c15112fddae7c4509ad639b4800a369c,
• f0ba941f62e7fdadd76dad4cd93c1893

lost their relations with unknown reports at all, while there is one new associated mark 94cda75e1c23421736e612ef4131c545 (normal CPU timeout).

This changes are because BenchExec allows to detect CPU timeouts in the way we need.

To fix new issues, please, see at the attached job archive.

Fixed.

Now everything works! I merged the branch to master in aa70aed85.