This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: glibc 2.18 freeze!


On 06/12/2013 04:29 PM, Adhemerval Zanella wrote:
> On 12-06-2013 17:15, Carlos O'Donell wrote:
>> On 06/12/2013 02:47 PM, Winfried Magerl wrote:
>>> gcc-4.8.1 is already released and even with a future gcc-fix (if bug is
>>> accpeted) it might be necessary to add some workaround for this problem
>>> befor releasing glibc-2.18.
>> Adhemerval,
>>
>> Given your familiarity with the issue do you have time to fix the 
>> memset issue in glibc so that we can compile with the already released 
>> gcc-4.8.1?
>>
>> One solution could be a configure check that detects the transformation
>> is present, sets a flag, and then disables the transformation when building
>> the appropriate file.
>>
>> The best example I can give is a recent fix I did for detecting when
>> __sync_val_compare_and_swap is used e.g. commit
>> a01f19c8fb12eef419d4112879bc715e2ab6f6d7
>>
>> You can use a variable to set something in config.make, which then
>> gets used to decided to enable or disable the feature.
>>
>> Cheers,
>> Carlos.
>>
> Hi Carlos,
> 
> I think I do, I'll setup a x86_64 environment and check your changes to
> __sync_val_campare_and_swap.

If __attribute__ ((optimize ("no-tree-loop-distribute-patterns"))) works
on older compilers with nor warning or error then that should also work.
I forgot we could use an attribute here.

c.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]