This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On Fri, 2009-07-31 at 03:41 +0100, Dave Korn wrote: > John David Anglin wrote: > > On Fri, 31 Jul 2009, Dave Korn wrote: > > > >> John David Anglin wrote: > >>> The c954a01 and c9a011b tests both fail somewhat randomly on more > >>> than one platform. I believe that the tests should be skipped. > >> I think they should be investigated first, shouldn't they? We could have a > >> real bug and be generating indeterminate code, couldn't we? > > > > c954a01 doesn't fail when not run from the testsuite on hppa. It > > also fails on powerpc. c9a011b fails intermittedly on hppa. Richard > > Guenther said it should be XFAILed because it also fails repeatedly > > for him. > > Ah, if it looks like an infrastructure problem that's a whole different kettle > of fish, makes sense to me now. I'll see if the same issue shows up on Cygwin, > I'm just in the course of fixing Ada problems for the platform anyway. Hi, A few monthes ago Jakub sent me a list of randomly failing tests from his x86_64-linux tester. c954a01 is in his list but not c9a011b. After some checking I found that the tests in question were failing mostly on multicore machines with low load, at the rate of about 1 out of 10000 (when test run in loop) and that it was likely related to timing issues, that is to say by changing some timing constants I was no longer able to get those tests to fail. I came up with the following patch tweaking timing constants for those tests, Jakub is curently testing it and I will formally submit it it improves things. Laurent
Attachment:
patch-c9-random-fail-v2.txt
Description: Text document
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |