[Python-Dev] Who can make test_fork1 fail? (original) (raw)

Tim Peters tim_one@email.msn.com
Sat, 19 Aug 2000 20:45:32 -0400

[Trent Mick, on //sourceforge.net/patch/?func=detailpatch&patch_id=101226&group_id=5470 ]

That patch seems to fix it for me. As before, I can get testfork to fail intermittently (i.e. it doesn't hang every time I run it) without the patch and cannot get it to hang at all with the patch.

Thanks a bunch, Trent! (That's a Minnesotaism -- maybe that's far enough North that it sounds natural to you, though .)

Would you like me to run and provide the instrumented output that I showed last time this topic came up?

Na, it's enough to know that the problem appears to have gone away, and since this was-- in some sense --the simplest of the test cases (just one fork), it provides the starkest contrast we're going to get between the behaviors people are seeing and my utter failure to account for them. OTOH, we knew the global lock should be a problem here (just not the problem we actually saw!), and Charles is doing the right kind of thing to make that go away.

I still encourage everyone to run all the tests that failed on all the SMP systems they can get hold of, before and after the patch. I'll talk with Guido about it too (the patch is still a bit too hacky to put out in the world with pride ).