Re: Question about some fields in regex's re_pattern_buffer

[Reuben Thomas]

On 15 August 2010 14:45, Paolo Bonzini <address@hidden> wrote:
>
> The non-thread-safety isn't only for RE_NO_SUB, but for all syntax options.
>  RE_NO_SUB makes it worse only because it is _not_ a syntax option, but a
> setting to provide improved matching speed.
>
> I think RE_NO_SUB should be documented, and a not on non-thread-safety of
> re_set_syntax_options should be added.

Agreed. (I presume you mean that a "note" should be added.)

However, this does raise a more general point, namely the wisdom of
spending effort maintaining the GNU API, rather than extending the
POSIX API where it is deficient with respect to the GNU API. The cost
of maintaining the GNU is relatively low, since so much of the code is
shared with the POSIX, and it is pehaps easier to fix the GNU API than
to extend the POSIX API (owing to the care needing to be taken in the
latter case not to break conformance). Nonetheless, far more
developers stand to benefit from, for example, adding backwards
searching to the POSIX API, than from adding thread-safety to the GNU
API.

Thoughts?

-- 
http://rrt.sc3d.org