|Summary:||icu fails in rawhide with gcc 4.7.0 mass rebuild|
|Product:||[Fedora] Fedora||Reporter:||Peter Robinson <pbrobinson>|
|Component:||gcc||Assignee:||Jakub Jelinek <jakub>|
|Status:||CLOSED RAWHIDE||QA Contact:||Fedora Extras Quality Assurance <extras-qa>|
|Version:||rawhide||CC:||denis.arnaud_fedora, erack, jakub|
|Fixed In Version:||gcc-4.7.0-0.9.fc17||Doc Type:||Bug Fix|
|Doc Text:||Story Points:||---|
|Last Closed:||2012-01-24 07:37:56 UTC||Type:||---|
|oVirt Team:||---||RHEL 7.3 requirements from Atomic Host:|
Description Peter Robinson 2012-01-20 15:21:12 UTC
icu 4.8.1 previously built fine in rawhide in the mass rebuild it failed in the tests with quite a number of test failures. Build is http://koji.fedoraproject.org/koji/taskinfo?taskID=3717531 The previous build on December 13th built OK so its possibly a regression in gcc http://koji.fedoraproject.org/koji/buildinfo?buildID=278730 I've tested the build up to and including gcc-4.7.0-0.8.fc17
Comment 1 Peter Robinson 2012-01-20 15:21:51 UTC
assigning to gcc as likely a gcc issue
Comment 2 Jakub Jelinek 2012-01-20 15:36:41 UTC
That needs analysis by the package maintainer first. It might be a GCC issue, but it might also be (and more probably) just package bug, strict aliasing violation, undefined behavior code etc. If you suspect a gcc bug, it needs to be first reduced into a small self-contained reproducer. Thanks.
Comment 3 Eike Rathke 2012-01-20 17:25:44 UTC
(In reply to comment #0) > icu 4.8.1 previously built fine @pbrobinson: I noticed you updated the ICU to version 18.104.22.168, did this package actually build with other gcc versions?
Comment 4 Peter Robinson 2012-01-21 14:01:20 UTC
(In reply to comment #3) > (In reply to comment #0) > > icu 4.8.1 previously built fine > > @pbrobinson: I noticed you updated the ICU to version 22.214.171.124, did this package > actually build with other gcc versions? Scratch build on f16 builds fine http://koji.fedoraproject.org/koji/taskinfo?taskID=3719806
Comment 5 Jakub Jelinek 2012-01-21 14:40:38 UTC
This turned to be a GCC bug, icu/source/test/cintltst/cbiditst.c u16ToPseudo function is miscompiled. But next time I'd really appreciate if at least the binary search (if doing brute force search) would be done by package maintainers who know their code better than me.