Red Hat Bugzilla – Bug 1012698
valgrind on ARM finds numerous glibc errors
Last modified: 2013-11-04 16:53:18 EST
Created attachment 803631 [details]
Failed stp build on ARM koji
Description of problem:
I thought we had the stp build fixed with the recent valgrind change, but, alas! That fixed the build on i686 and x86_64 only. Now the ARM build is failing, with a long list of errors found inside glibc routines. Since the koji build logs will eventually disappear, I am attaching the entire build log to this bug report.
See http://koji.fedoraproject.org/koji/taskinfo?taskID=5989445 for as long as that URL is good.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. fedpkg clone stp
2. cd stp
3. fedpkg srpm
4. Build the source RPM on ARM
The attached build log shows the stp build failing its tests due to errors reported by valgrind inside glibc routines.
No errors (unless there really is an stp bug on ARM, of course).
Was there ever a successful build on arm before? I cannot quickly find one in koji and the old arm.koji doesn't seem to use valgrind?
Analysing this might take some time. valgrind ARM support is less mature than other arches. But a lot of ARM work has been done for the upcoming 3.9.0 version.
You're right. We weren't doing arm builds before F-20, so this is new. I have disabled the valgrind tests on arm for now. I will try them again once in awhile after I see a valgrind update to see if they can be re-enabled. Thanks again.
I did some scratch builds against current upstream valgrind svn:
Results also posted upstream here:
Thanks for the pointers. I'll follow the upstream discussion.
valgrind 3.9.0 is now in updates-testing for f20 and in rawhide. Would you be able to test against that version?
Hey, nice, that solves the problem. Valgrind is now silent except for when there's an actual problem in the stp code. Thanks a ton, Mark.
Thanks for checking.