Threads, glibc, and all that

[Mark Mitchell]

I believe that the recent changes that HJ Lu and Jason Merrill have
collaborated on, and which are (I believe) in 1.0.2, seem to be
causing trouble on my i686-pc-linux-gnu (RedHat 5.0) system.

In partictular, the g++.brendan/crash30.C test fails:

    In file included from /home/mitchell/projects/egcs/libstdc++/stl/stl_alloc.h:73,
		     from /home/mitchell/projects/egcs/libstdc++/stl/alloc.h:21,
		     from /home/mitchell/projects/egcs/libstdc++/std/bastring.h:39,
		     from /home/mitchell/projects/egcs/libstdc++/string:6,
		     from crash30.C:3:
    /usr/include/pthread.h:438: parse error before `*'

There seems to be a problem with sigset_t not being defined, since
that line in pthread.h looks like

extern int pthread_sigmask __P ((int __how, __const sigset_t *__newmask,
				 sigset_t *__oldmask));

The crash30.C test is just:

    #include <string>

    main(void) {

      string a[] = {"Hello"};

    }

The test only fails when compiled -ansi, as is done in the testsuite.
 
-- 
Mark Mitchell <mmitchell@usa.net>
http://home.earthlink.net/~mbmitchell
Consulting Services Available