Bug 137440 - mdadm uses funky pointer dereferencing which breaks aliasing rules
|
Bug#:
137440
|
Product: Gentoo Linux
|
Version: 2006.0
|
Platform: All
|
|
OS/Version: Linux
|
Status: RESOLVED
|
Severity: normal
|
Priority: P2
|
|
Resolution: FIXED
|
Assigned To: base-system@gentoo.org
|
Reported By: agaffney@gentoo.org
|
|
Component: Applications
|
|
|
URL:
|
|
Summary: mdadm uses funky pointer dereferencing which breaks aliasing rules
|
|
Keywords:
|
|
Status Whiteboard:
|
|
Opened: 2006-06-20 15:35 0000
|
The summary says it all. mdadm-1.12.0 builds fine with gcc-3.4.6 and fails in
gcc-4.1.1 due to warnings being treated as errors in gcc-4.1.1. The latest ~x86
(2.5) builds just fine with both gcc versions. Perhaps we can stable a newer
version in time for the release snapshot?
This only happens when you add '-O2' to CXFLAGS, as done by the ebuild.
I suggest to 'filter-flags "-O*"'.
vapier: this is a potential release blocker for amd64, so I'd be delighted to
have it rather solved before the snapshot date.
what does that have to do with adding a hack like filtering flags ? fix it
properly or dont fix it at all
(In reply to comment #5)
> what does that have to do with adding a hack like filtering flags ? fix it
> properly or dont fix it at all
I don't think it was clear what your previous comment was regarding. I took it
the same way that Kugelfang did...that the bug itself was bogus.
the bug is not bogus, later versions of mdadm just dont compile with -Werror
see the recent thread on gentoo-dev mailing list about aliasing rules
No, no, no...I know the bug isn't bogus. I wouldn't have filed it :P I was
saying that Kugelfang (probably) and I took your comment (#3) to mean that you
thought the bug was bogus.
2.5-r1 now has a patch to fix this properly
Since the current stable is nowhere near 2.5-r1, can we get this reopened and
assigned to the proper arch teams for stabilization? Fixing it in the tree
doesn't help Release Engineering if it isn't stable, too.