Summary: | dev-lang/ruby-1.9.* - segfault in test_fork.rb on HPPA | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Jeroen Roovers (RETIRED) <jer> |
Component: | Current packages | Assignee: | Gentoo Ruby Team <ruby> |
Status: | RESOLVED NEEDINFO | ||
Severity: | normal | CC: | hppa, sandbox |
Priority: | Normal | Keywords: | TESTFAILURE |
Version: | unspecified | ||
Hardware: | HPPA | ||
OS: | Linux | ||
URL: | http://redmine.ruby-lang.org/issues/5476 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | 342819 | ||
Bug Blocks: | |||
Attachments: | dev-lang:ruby-1.9.3_rc1:20111022-170259.log |
Description
Jeroen Roovers (RETIRED)
2011-10-22 17:22:32 UTC
Filed a bug upstream about it. Just noticed a comment on the upstream bug report that ruby on hppa is broken because you have linuxthreads, not nptl. I've also added 1.9.3p0 to the tree. You might want to try although there is nothing that indicates that this has been fixed. That's rubbish. HPPA has had NPTL for years, or at least Gentoo/HPPA. AFAIK Debian never did the NPTL port, but we did. Jeroen, could you get a backtrace for this segfault? I don't think this bug is going anywhere upstream without it. 1.9.3_p0 exhibits the same segmentation fault. Interestingly, when I set `ulimit -c unlimited' in the build directory and run `make -j1 test', the segmentation fault does not occur. The test suite is also fine when the maximum stack size is unlimited (ulimit -s unlimited), which is probably why `ulimit -c unlimited' works too. I just checked the upstream bug and there has not been any movement since they are waiting for more information. It would be great if you could test this with the latest in-tree version, since there have been a lot of bug fixes and this might just be fixed via another bug. Otherwise I guess we'll have to try and get upstream the information they want. (In reply to comment #8) > http://bugs.ruby-lang.org/issues/5488 Ah no, this belongs to bug #342819. Fixed in >=dev-lang/ruby-1.9.3_p194-r1. This still happens with newer versions but only when FEATURES=sandbox, so I assume there is a HPPA specific (stack related?) problem there. Does it still happen with recent ruby versions? What about ruby20 / ruby21? Yes, it seems to have been fixed somewhere at some point, possibly in the HPPA linux kernel. I can't reproduce it now. |