Bug 1536431 - redhat-rpm-config: Be more explicit about -fasynchronous-unwind-tables
Summary: redhat-rpm-config: Be more explicit about -fasynchronous-unwind-tables
Keywords:
Status: CLOSED RAWHIDE
Alias: None
Product: Fedora
Classification: Fedora
Component: redhat-rpm-config
Version: rawhide
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Florian Weimer
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2018-01-19 12:23 UTC by Florian Weimer
Modified: 2018-03-02 09:28 UTC (History)
11 users (show)

Fixed In Version: redhat-rpm-config-82-1.fc28
Clone Of:
Environment:
Last Closed: 2018-01-22 15:03:02 UTC
Type: Bug
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Bugzilla 1548847 0 unspecified CLOSED gcc: Generate unwind tables by default on ppc64, ppc64le 2021-02-22 00:41:40 UTC
Red Hat Bugzilla 1550914 0 unspecified CLOSED glibc: Missing unwind tables on ppc64, ppc64le 2021-02-22 00:41:40 UTC

Internal Links: 1548847 1550914

Description Florian Weimer 2018-01-19 12:23:21 UTC
Upstream GCC enables the flag on i686, x86-64, s390x, and we patch it in for aarch64.  ppc64, ppc64le, armv7hl/armhfp should use it as well.

The flag is needed so that performance/debugging tooling works.  glibc cancellation support needs it as well under certain circumstances.

Comment 1 Jakub Jelinek 2018-01-19 12:26:16 UTC
Unsure about armv7hl/armhfp, that doesn't use DWARF unwind info, but some EABI stuff, no idea what this option does in that case, maybe nothing, somebody would need to test that.

Comment 2 Florian Weimer 2018-01-22 14:16:47 UTC
ppc64, ppc64le, armv7hl/armhfp do not need it due to differences in stack management.  Updating summary.

Comment 3 Mark Wielaard 2018-02-25 13:10:41 UTC
(In reply to Florian Weimer from comment #2)
> ppc64, ppc64le, armv7hl/armhfp do not need it due to differences in stack
> management.  Updating summary.

Disabling this for ppc64/ppc64le (in glibc) seems to have caused a regression causing the elfutils testsuite to be unable to unwind in some cases because of missing CFI in the .eh_frame (CFI only being available through .debug_frame in the external debuginfo package) for glibc functions like raise. See e.g. https://koji.fedoraproject.org/koji/taskinfo?taskID=23812571

Comment 4 Mark Wielaard 2018-02-25 13:12:14 UTC
(In reply to Mark Wielaard from comment #3)
> (In reply to Florian Weimer from comment #2)
> > ppc64, ppc64le, armv7hl/armhfp do not need it due to differences in stack
> > management.  Updating summary.
> 
> Disabling this for ppc64/ppc64le (in glibc) seems to have caused a
> regression causing the elfutils testsuite to be unable to unwind in some
> cases because of missing CFI in the .eh_frame (CFI only being available
> through .debug_frame in the external debuginfo package) for glibc functions
> like raise. See e.g.
> https://koji.fedoraproject.org/koji/taskinfo?taskID=23812571

Note that this might actually be glibc specific, since this happens on recent f27 too.

Comment 5 Florian Weimer 2018-02-25 13:20:50 UTC
I filed bug 1548847 against Fedora gcc.  The expectation is that redhat-rpm-config follows the GCC defaults in this area.


Note You need to log in before you can comment on or make changes to this bug.