Bug 1386126
Summary: | binutils-2.27-7.fc26 and later break Haskell dynamic linking on aarch64 and armv7hl | ||
---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | Jens Petersen <petersen> |
Component: | binutils | Assignee: | Nick Clifton <nickc> |
Status: | CLOSED RAWHIDE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
Severity: | unspecified | Docs Contact: | |
Priority: | unspecified | ||
Version: | rawhide | CC: | jakub, nickc, petersen |
Target Milestone: | --- | ||
Target Release: | --- | ||
Hardware: | arm | ||
OS: | Unspecified | ||
Whiteboard: | |||
Fixed In Version: | binutils-2.27-10.fc26 | Doc Type: | If docs needed, set a value |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2016-11-25 02:11:06 UTC | Type: | Bug |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: |
Description
Jens Petersen
2016-10-18 08:46:01 UTC
Ah to reproduce this you shuold use F25 together with F26 binutils since i added a workaround to F26 ghc-rpm-macros for this. Alternatively you can use Rawhide together with f25 ghc-rpm-macros (or just delete the final "%ghc_without_dynamic" line in "/usr/lib/rpm/macros.d/macros.ghc".) Hi Jens, Please may I ask a few questions first, before I attempt to reproduce this problem: * Does the problem exist with other architectures, eg x86_64, or just ARM and AArch64 ? (I assume the latter, but if this is a sysroot problem then it is suspicious that only those two architectures are affected). * Is it only the logf symbol that cannot be resolved, or are there (lots of) other symbols affected as well ? If it is just logf then this would might indicate a problem with the dynamic library involved (libm.so ?) rather than the linker. Cheers Nick Hi Nick, Yes it just happens on armv7hl and aarch64. From memory I think affects various (all?) symbols starting with "log" but I would need to check the logs more carefully to be sure. I am pretty sure it is not just "logf". Here are some examples: http://koji.fedoraproject.org/koji/taskinfo?taskID=16103782 http://koji.fedoraproject.org/koji/taskinfo?taskID=16103673 Linking dist/build/xmobar/xmobar ... /usr/lib/ghc-7.10.3/scien_3vmCXiZeJinKZBcLawS8i0/libHSscientific-0.3.4.7-3vmCXiZeJinKZBcLawS8i0-ghc7.10.3.so: error: undefined reference to 'log' http://koji.fedoraproject.org/koji/taskinfo?taskID=16071865 Linking dist/build/darcs/darcs ... /builddir/build/BUILD/darcs-2.12.2/darcs-2.12.2/dist/build/libHSdarcs-2.12.2-5w02tFeyqS4LTTczMtN7Ff-ghc7.10.3.so: error: undefined reference to 'log' /usr/lib64/ghc-7.10.3/text_HmqVQnZSpjaC156ABqPhne/libHStext-1.2.2.1-HmqVQnZSpjaC156ABqPhne-ghc7.10.3.so: error: undefined reference to 'logf' (older logs are gone from koji) I agree it is suspicious about libm.so. However I could reproduce the problem directly on armv7hl F25 by installing binutils from rawhide there, so I am not sure it is a glibc issue. Hi Jens, OK - So I can now reproduce the problem in a VM. (Boy is it slow...) I have found a workaround - use the BFD linker instead of the GOLD linker - but I have not yet found the source of the problem. Part of the issue is that it is proving very hard to capture the full linker command line that is being used. I have a hack that seems to be working for the moment though, so I can proceed. Cheers Nick Hi Jens, I believe that I have found a fix for this problem. Please could you try out: binutils-2.27-10.fc26.rpm and let me know how you get on. Cheers Nick Sorry I missed your comments - I was traveling at the beginning of the month. I just did a scratch build of cpphs in Rawhide (binutils-2.27-12.fc26) without %ghc_without_dynamic and it seemed to succeed: http://koji.fedoraproject.org/koji/taskinfo?taskID=16607727 so your fix looks good to me. I will remove the workaround from ghc-rpm-macros. :-) Thank you very much and sorry for the slow feedback. *** Bug 1385475 has been marked as a duplicate of this bug. *** *** Bug 1385477 has been marked as a duplicate of this bug. *** |