Problem with gcc4.7 and call_once

[David Barto]

So why does building the code from scratch with the --enable-libstdcxx-threads flag work, and not when using the port or when the port file configuration adds that option?

Jeremy has shown that the MP version does not work for std::call_once, and gcc.org has results showing that it works fine when built and tested by them (bug closed at that site now).

That appears to be the real issue.

	David

On Aug 9, 2013, at 9:02 AM, Jeremy Huddleston Sequoia <jeremyhu at macports.org> wrote:

> 
> On Aug 9, 2013, at 8:23, Brian D. McGrew <brian at visionpro.com> wrote:
> 
>> One more questionŠ  you know a lot more about port than I so I have a no
>> reason to doubt when you say it's on by default.  But in the port
>> installed versions when I do a gcc-mp -v, I do not see
>> --enable-libstdcxx-threads in the configured line.  When I do the same on
>> my hand cranked version, I do see the option on the configured line.
> 
> Yes.  gcc -v just reports back the configure options passed in.  Ones that are on by default are not printed unless you actually specified them.
> 
> --Jeremy
> 
>> 
>> -b
>> 
>> 
>> 
>> On 8/9/13 8:17 AM, "Jeremy Huddleston Sequoia" <jeremyhu at macports.org>
>> wrote:
>> 
>>> 
>>> On Aug 9, 2013, at 6:27, Brian D. McGrew <brian at visionpro.com> wrote:
>>> 
>>>> Well I could be going in the wrong direction, I'm not the foremost
>>>> expert
>>>> on ports yet, but what I do know is this:
>>>> 
>>>> If I port edit gcc48 and add --enable-libstdcxx-threads and then port
>>>> install -s gcc48, when I run gcc -v I do see the option that I just
>>>> added.
>>>> No question, the option IS there.  But the test program still doesn't
>>>> run.
>>> 
>>> 
>>> Yes, I would expect there to be no change in behavior since
>>> --enable-libstdcxx-threads is on by default (thus your adding it has no
>>> effect other than to show it in 'gcc -v'
>>> 
>>>> If I rebuild gcc-4.8.1 from source by hand using the same configuration
>>>> options from gcc-mp -v and add --enable-libstdcxx-threads at
>>>> configuration
>>>> time and do a make install over top of the existing gcc-mp, everything
>>>> works fine.  
>>>> 
>>>> Now I'm really confused.
>>> 
>>> Yes, something is different between your out-of-tree-build and your
>>> in-tree build.  I'll try to repro your working case and then bisect the
>>> two.
>>> 
>>> It would be great if someone without GPL3phobia could look at this.  It's
>>> probably rather simple to triage with source access, but at least having
>>> a working and broken build configurations from the same sources will let
>>> me bisect the build configurations.
>>> 
>>> Thanks,
>>> Jeremy
>>> 
>> 
>