Re: libpthread in glibc (was: Bug#407208: imagemagick: convert hangs on hurd-i386)

[Neal H. Walfield]

At Sun, 4 Feb 2007 13:38:15 +0100,
Thomas Schwinge wrote:
> On Wed, Jan 31, 2007 at 03:28:26PM +0100, Samuel Thibault wrote:
> > Samuel Thibault, le Wed 31 Jan 2007 00:32:17 +0100, a ?crit :
> > > Unfortunately this is not enough: oddly enough, libX11's weak definition
> > > wins (at least in glibc 2.3.6 and 2.5.0, but not in glibc 2.2.5)...  I
> > > fail to see how weak definitions may be useful if strong definitions
> > > can't compete. I submitted a bug report to glibc's bugzilla.
> > 
> > Answer from Drepper: this "feature" was fixed (understand: removed) in
> > glibc 2.3, not only because it is bit more conforming, but also because
> > it was needed for some other stuff.
> > 
> > As a result, our only solution is to at last implement stubs in libc for
> > pthread_mutex_lock & co.
> 
> How much work would be needed to plug the libpthread into glibc?  Is only
> build system work needed and having the initialization code called at the
> right places or are there more subtle problems to be expected?

That should be about it.  There is some possible integration work,
e.g., pthread_atfork, but that would already be improving on the
status quo.