Re: building glibc 2.15

[Allan McRae <mcrae_allan at hotmail dot com>]

On 23/03/12 02:46, Allin Cottrell wrote:
> On Thu, 22 Mar 2012, Thomas Schwinge wrote:
> 
>> On Thu, 22 Mar 2012 08:42:41 -0400 (EDT), Allin Cottrell
>> <cottrell@wfu.edu> wrote:
>>> I just tried building glibc 2.15 from the release tarball on
>>> i686-pc-linux-gnu, and I see that the old __i686 issue is
>>> still there (build failure at i386/fpu/s_frexp.S).
>>>
>>> I got the impression from reading libc-alpha that this (and
>>> various other things) had been fixed lately. Could anyone tell
>>> me if the __i686 thing is in fact fixed in git, and if so
>>> what's the recommended git invocation to get the latest
>>> "goodies".
>>
>> Please see <http://sourceware.org/glibc/wiki/Release/2.15#Build_Failures>
>> for the __i686 issue, and <http://sourceware.org/glibc/wiki/GlibcGit> for
>> Git instructions.
> 
> Thanks. After I applied the __i686 patch found on the wiki glibc 2.15
> built OK. But I'm seeing a few failures on "make check" and wondered if
> they were "expected" or anomalous. The failed tests are tst-mqueue1,
> tst-mqueue2, tst-mqueue3, tst-mqueue4, tst-mqueue7 and check-localplt. I
> can post the .out files if that's useful, but for example:
> 
> "mq_setattr on invalid mqd_t did not fail with EBADF: \
> Function not implemented"
> 
> "fct: mq_getattr failed: Function not implemented"
> 
> This is on Linux 3.3, gcc 4.7.0.


I am in the progress of upgrading my toolchain on i686 and also got
these tst-mqueue* errors.  The update involved:

gcc-4.6.3 -> 4.7.0
linux-headers-3.1.6 -> 3.3
binutils-2.22.0.20111227 -> 2.22.0.20120323

with glibc untouched. I will investigate further (firstly limiting it to
a specific component of the upgrade)

Allan