mingw-cross-env-list
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Mingw-cross-env-list] Solving the Pthreads issue


From: Tony Theodore
Subject: Re: [Mingw-cross-env-list] Solving the Pthreads issue
Date: Thu, 20 May 2010 17:28:28 +1000

On 20 May 2010 09:41, Volker Grabsch <address@hidden> wrote:
> Tony Theodore <address@hidden> schrieb:
>> On 17 May 2010 02:06, Volker Grabsch <address@hidden> wrote:
>> > I performed some adjustments and included it. Thanks!
>> > http://hg.savannah.gnu.org/hgweb/mingw-cross-env/rev/a13b50a30182
>> >
>> > Tony, are you able to reproduce the threading issue with this
>> > new script?
>> >
>> > I.e., is it a bug in pthreads-win32 or in libgomp?
>>
>> I couldn't reproduce the problem with pthreads, so I imagine it's an
>> issue with either the autostatic patch or libgomp. I haven't had a
>> chance yet to try isolate it further, but I suspect it's the
>> autostatic patch since libgomp is fine with a shared build.
>
> I just improved the pthreads/libgomp examples and included
> the autostatic patch.
> http://hg.savannah.gnu.org/hgweb/mingw-cross-env/rev/c206341fa616
> http://hg.savannah.gnu.org/hgweb/mingw-cross-env/rev/2429543e1aa9
>
> I'm still observing some strange behaviour, i.e. libgomp
> seems to use always the same thread for everything. The
> pthreads test program, however, works fine.

That's pretty much the same as my observations, though there are a few
other notes:

The autostatic patch causes the vmime build to fail, but it's order
dependent. A make clean && make vmime will work, but the normal order
will build pthreads before vmime. I imagine this is due to
--without-pthread being an unrecognized option in vmime, and configure
detects it as both "no" and a little later "yes".

The autostatic patch is supposedly against pthreads-cvs, even though
it applies and builds cleanly without the cvs update patch. The cvs
update patch however, causes the wxwidgets build to fail (but I
haven't looked into it).

The autostatic patch conflicts with the win64 patches - not a current
issue, just a note.

> Question to all:
>
> Can with live with that, at least for the next release?
> What do you think?

Pthreads support has come a long way since the last release, and the
remaining issues are more easily isolated. Nothing depends on libgomp,
and vmime should be `just` some configure tweaks. Now would be a good
time for release while almost everything else is relatively stable.

Tony



reply via email to

[Prev in Thread] Current Thread [Next in Thread]