This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [s390] Build failure with gcc 4.9.2 in rawhide
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: Joseph Myers <joseph at codesourcery dot com>, Andreas Krebbel <krebbel at linux dot vnet dot ibm dot com>
- Cc: Siddhesh Poyarekar <siddhesh at redhat dot com>, libc-alpha at sourceware dot org
- Date: Mon, 26 Jan 2015 23:38:10 -0500
- Subject: Re: [s390] Build failure with gcc 4.9.2 in rawhide
- Authentication-results: sourceware.org; auth=none
- References: <20150123131340 dot 1a703d6e at redhat dot com> <54C6530E dot 2060601 at linux dot vnet dot ibm dot com> <alpine dot DEB dot 2 dot 10 dot 1501261818130 dot 28603 at digraph dot polyomino dot org dot uk>
On 01/26/2015 01:20 PM, Joseph Myers wrote:
> On Mon, 26 Jan 2015, Andreas Krebbel wrote:
>
>> Hi Siddhesh,
>>
>> On 01/23/2015 08:43 AM, Siddhesh Poyarekar wrote:
>>> Hi Andreas,
>>>
>>> We encountered a build failure in rawhide for s390x due to -Werror:
>>>
>>> http://s390.koji.fedoraproject.org/koji/getfile?taskID=1705688&name=build.log&offset=-24000
>>>
>>> I was able to narrow it to a small reproducer and have filed it as a
>>> bug in gcc:
>>>
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64739
>>>
>>> I have unbroken the build temporarily in Fedora by disabling werror
>>> for s390x. Could you please have this patched in glibc before the
>>> release?
>>
>> the difference between S/390 and common code (gconv_simple.c) appears to
>> be that we define STORE_REST and gconv_simple.c doesn't. This triggers
>> the GCC bug. Your reduced testcase fails on x86 as well. When I replace
>> the !STORE_REST code in gconv_simple.c with the STORE_REST code from
>> loop.c I get the same error message on x86. So to my understanding there
>> is nothing to do apart from fixing GCC - right?!
>
> Well, if the warning appears with GCC 4.9, then suppressing with DIAG_*
> macros (conditional on STORE_REST), naming 4.9 as the relevant version,
> would seem appropriate - the comment should reference the GCC bug, so
> making it easy to tell at what point the suppression can be removed (e.g.
> when moving to GCC 5 as minimum version, if the bug gets fixed in GCC 5).
Agreed.
Please feel free to check this in immediately to 2.21.
Cheers,
Carlos.