This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 10/17] Regex: Portability to non-GCC compilers.
On 08/12/2017 09:04, Florian Weimer wrote:
> On 12/08/2017 10:47 AM, arnold@skeeve.com wrote:
>> Florian Weimer <fweimer@redhat.com> wrote:
>>
>>> On 12/08/2017 10:16 AM, Arnold Robbins wrote:
>>>> +#if !defined(__GNUC__) || __GNUC__ < 3
>>>> + static short utf8_sb_map_inited = 0;
>>>> +
>>>> + if (! utf8_sb_map_inited)
>>>> + {
>>>> + int i;
>>>> +
>>>> + utf8_sb_map_inited = 0;
>>>> + for (i = 0; i <= 0x80 / BITSET_WORD_BITS - 1; i++)
>>>> + utf8_sb_map[i] = BITSET_WORD_MAX;
>>>> + }
>>>> +#endif
>>>
>>> This doesn't look like a good idea because it's not thread-safe.
>>
>> Is the rest of regex thread safe?
>>
>> I've no objection to something else that does the trick for GLIBC.
>> Gawk doesn't have multiple threads.
>
> It does not matter for glibc because it's compiled with GCC. It might matter if this code is merged into gnulib, which is supposed to be portable to many compilers and environments.
My understanding is we still aim to keep in sync with gnulib, so I think
we should first integrate with current gnulib code adding any glibc code
(if any and with the idea of integrating it back to gnulib). I do not see
much point in deviate even more from gnulib, unless the idea is to really
decouple both implementations.