PATCH: Provide runtime intialization of _STL_mutex_lock class (revised)

[John David Anglin]

> Assuming loren has no objections, and there are no regression with it,  
> it's ok with me for both trunk and branch. I'd prefer if no warnings were 
> issued, and the slight tweaks to gthr were made.

The warning only occurs for the __GTHREAD_MUTEX_INIT_FUNCTION case.
The problem with the tweak to the GTHREAD paradigm is that changes
need to be made to the the included gthr files that define
__GTHREAD_MUTEX_INIT_FUNCTION.  You need to know the details of
the __ghtread_mutex_t type to define the initializer.  I can easily
do this for gthr-dce.h.  I don't have the information needed for
the others.  It's easy to check whether this initializer is defined,
so if you like this idea I could modify the patch to use say
__GTHREAD_MUTEX_INIT_DEFAULT to complete the initialization.
Separate patches to gthr.h and gthr-dce.h could then be developed
to document and define this initializer for hpux.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)