Re: [Gnash-dev] configure malfunction detecting pthreads with configure

[Eric Hughes]

At 01:32 PM 10/28/2007, Bastiaan Jacques wrote:
> -  *mingw*)
> +  *mingw* | *cygwin*)
>      pthread_flags="-mthreads"

This works insofar as it goes.  It does get rid of the annoying -Kthread 
"unknown option" messages.  Thanks, and committed.

There are other problems, however.  The problem is with the following lines 
in pthreads.m4:
>   for i in $libslist; do
>     if test -f $i/libpthread.a -o -f $i/libpthread.${shlibext} -o -f 
> $i/libpthread.dylib; then
>       pthreads=yes

Cygwin has /usr/lib/libpthread.a (that's on the cygwin file system), but 
it's not a MinGW library, it's the cygwin one.  MinGW does not come with 
pthreads by default and I haven't installed it.  What does seem clear is 
that $libslist is a likely culprit, rather than anything particularly 
specific in pthreads.m4.  I don't know enough about the libraries and their 
layout to know what the right and good solution is.

The larger issue is that in an ideal world the whole pthreads test could be 
skipped entirely in this instance.  I understand from Rob that it's there 
to ensure that the Boost Thread library links correctly, as it uses 
pthreads on *x platforms for its implementation there, but not for all 
platforms.  From what I can tell at this moment, Boost.Thread on MinGW uses 
native Win32 threads *instead* of pthreads (as does MSVC), but I'm working 
with inadequate documentation.  All gnash dependencies on pthreads were 
replaced with Boost threads a few months ago, so pthreads has become a 
conditional dependency in the interim.

Eric