[bug-diffutils] bug#24228: bug#24228: [platform-testers] new snapshot available: diffutils-3.4.9-d647

[Jim Meyering]

On Thu, Sep 8, 2016 at 5:02 AM, Dagobert Michelsen <address@hidden> wrote:
> Hi Jim,
>
> Am 19.08.2016 um 06:18 schrieb Jim Meyering <address@hidden>:
>> On Mon, Aug 15, 2016 at 1:35 PM, Dagobert Michelsen <address@hidden> wrote:
>>>> Am 15.08.2016 um 07:33 schrieb Jim Meyering <address@hidden>:
>> ...
>>>> http://meyering.net/diff/diffutils-3.4.9-d647.tar.xz
>>
>> Thank you for testing.
>>
>>> I get this error on Solaris 10 Sparc with GCC 5.2, same for
>>> Solaris 11 Sparc with GCC 5.2:
>>>
>>>>  CC       rawmemchr.o
>>>> rawmemchr.c: In function 'rawmemchr':
>>>> rawmemchr.c:49:18: error: cast increases required alignment of target type 
>>>> [-Werror=cast-align]
>>>>   longword_ptr = (const longword *) char_ptr;
>>>>                  ^
>>>> cc1: all warnings being treated as errors
>>>> gmake[2]: *** [Makefile:1784: rawmemchr.o] Error 255
>>
>> In this case, I think the best advice it not to with warnings enabled
>> (don't use configure's --enable-gcc-warnings) unless you're using a
>> sufficiently modern system, probably glibc-based.
>
> Ok.
>
>>> Compilation on Solaris 11 x86, 32 and 64 bit work fine with
>>> Oracle Studio 12.4.
>>>
>>> On Solaris 11 Sparc with Oracle Studio 12.4 it works for 32 bit,
>>> for 64 bit I get an error in the testsuite:
>>>
>>>> FAIL: test-timespec
>>>> ===================
>>>>
>>>> test-timespec.c:138: assertion 'cmp (a, b) == sign (i - j)' failed
>>>> FAIL test-timespec (exit status: 262)
>>
>> This one is interesting.
>> It would be great if you could get in with a debugger to see what
>> exactly is going wrong.
>> It's probably not an issue for diffutils, as it barely uses timespec
>> at all and does not use timespec_sub.
>
> I didn’t get into looking into that yet, sorry.
>
>>> On Solaris 9 Sparc with GCC 4.6 I still get this, probably because the
>>> change you proposed in not in the release tarball. I’d happily test if the
>>> updated tarballs are bootstrapped:
>>>
>>>>  CC       error.o
>>>> error.c:112:1: error: function declaration isn't a prototype 
>>>> [-Werror=strict-prototypes]
>>
>> This appears to be exposing some unusually old code in error.c that
>> should instead be relying on gnulib's own strerror_r module.
>> We'll probably fix it properly in the long run, but simply not using
>> --enable-gcc-warnings will avoid it for you now.
>
> Ok.
>
>>> I would like to set up buildbot, to which repository? The one at Savannah?
>>> Do you have a preferred branch?
>>>  git://git.savannah.gnu.org/diffutils.git
>>
>> A buildbot would be great!
>> That is indeed the right repository, and as for branch, please use "master".
>
> This is now set up:
>   https://buildfarm.opencsw.org/buildbot/waterfall?tag=diffutils
>
> I will add more build options, compilers and platforms sometime.
>
> I can also configure sending automatic emails if a build breaks, which 
> destination
> address would you prefer?

Great!
Will get back to you shortly with an address. It'll probably be a new
mailing list.