- User Since
- Aug 18 2018, 5:33 PM (56 w, 14 h)
Oct 16 2018
Yes. Sorry I'm still not familiar with the tools here.
Aug 28 2018
Thank you for your comments everybody. I'll try to explain my motivations a bit and address the points raised. First of all my aim here is mainly readability, not performance. I found the current testsuite driver somewhat difficult to follow and the existence and mode of execution of the timeout scripts seemed like a hack that was no longer necessary as the python subprocess module had become capable of dealing with timeouts, process groups, etc. directly. Unfortunately it turned out that on Windows the process groups handling of subprocess, while functional on light loads, broke down quite quickly with heavier loads (as Phyx suspected on Trac). Nevertheless l thought it would still make the testsuite driver more readable if we would get rid of the timeout scripts, only this time rewriting the Windows part of timeout.hs functionality (the POSIX part is dead code at the moment as far as I see) as a python script that could be explicitly imported by the driver for Windows runs, instead of juggling separate executables through make and bash magic. I am also not completely convinced that this would be worth much, but to see how it would look like and what complications would arise I thought the easiest way was to just get up and write the patch.
Aug 26 2018
- 15363 Remember to send EOF after the input is written
Aug 25 2018
Fix a last moment edit that went on the wrong line
Include the first commit of the intended range, as it was accidently left out