On Wed, 25 Jan 2006 23:59:25 -0500, glyph@divmod.com wrote:

I just wanted to let everyone know that the recent buildbot failures on win32 (the process-related ones) are about half my fault. At the time that I wrote the offending code, I couldn't get a full test run for the win32 reactor due to a problem which James Knight has since fixed (THANK YOU) and the process tests were passing.

I already have an issue in my queue which describes the problem <http://twistedmatrix.com/bugs/issue654> and I should be fixing it at the latest this weekend.

Anybody know how to fix the other half (mostly _setUpTestdir)? The problems log for a recent win32select build is here: <http://twistedmatrix.com/buildbot/win32-select/builds/173/step-default/2>

There also seem to be some timing related issues which seem like they should just be todo'd until someone writes a version that has a fake timer rather than a real clock in them.

If you're referring to the newpb failures, I think we need some other way to deal with them. I've been trying to catch Warner online, but haven't succeeded so far. Should this code go back into a branch? Should buildbot just not run the newpb tests? Neither of these seems really satisfactory to me, but these failures shouldn't be failing builds: they never passed reliably in the first place. Jean-Paul

On Wed, 25 Jan 2006 23:59:25 -0500, glyph@divmod.com wrote:

I just wanted to let everyone know that the recent buildbot failures on win32 (the process-related ones) are about half my fault. At the time that I wrote the offending code, I couldn't get a full test run for the win32 reactor due to a problem which James Knight has since fixed (THANK YOU) and the process tests were passing.

I already have an issue in my queue which describes the problem <http://twistedmatrix.com/bugs/issue654> and I should be fixing it at the latest this weekend.

Anybody know how to fix the other half (mostly _setUpTestdir)? The problems log for a recent win32select build is here: <http://twistedmatrix.com/buildbot/win32-select/builds/173/step-default/2>

There also seem to be some timing related issues which seem like they should just be todo'd until someone writes a version that has a fake timer rather than a real clock in them.

If you're referring to the newpb failures, I think we need some other way to deal with them. I've been trying to catch Warner online, but haven't succeeded so far. Should this code go back into a branch? Should buildbot just not run the newpb tests? Neither of these seems really satisfactory to me, but these failures shouldn't be failing builds: they never passed reliably in the first place. Jean-Paul