[PATCH] Async signal safe TLS accesses

[Paul Pluzhnikov]

Joseph,

On Tue, Oct 22, 2013 at 2:47 PM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> On Tue, 22 Oct 2013, Andrew Hunter wrote:
>
>> +  INTERNAL_SYSCALL_DECL (err);
>> +
>> +  ret = INTERNAL_SYSCALL (rt_sigprocmask, err, 4, SIG_SETMASK, &new, old,
>> +                       _NSIG / 8);
>
> My comment from 3 Oct
> <https://sourceware.org/ml/libc-alpha/2013-10/msg00131.html> about
> syscalls not being usable directly in elf/, still applies.

Could you please provide some guidance on what needs to be done for Hurd?

Neither Andrew, nor I, know anything about Hurd signals.

Would it be ok to put a "stub" _dl_mask_all_signals() into
sysdeps/mach/hurd/dl-sysdep.c, and provide real implementation in
sysdeps/unix/sysv/linux/dl-sysdep.c ?

>> diff --git a/nptl/Makefile b/nptl/Makefile
>> index cd601e5..af63800 100644
>> --- a/nptl/Makefile
>> +++ b/nptl/Makefile
>> @@ -1,3 +1,4 @@
>> +
>>  # Copyright (C) 2002-2013 Free Software Foundation, Inc.
>>  # This file is part of the GNU C Library.
>>
>
> My comment about avoiding spurious changes applies here.

I've corrected this, and all other formatting deficiencies of Andrew's
patch (all that I could find, that is), and will resend it as soon as
I understand what to do about Hurd.

Thanks,

-- 
Paul Pluzhnikov