Re: new module 'sigprocmask'

[Eric Blake]

Bruno Haible <bruno <at> clisp.org> writes:

> 
> > I'm not a Windows expert, so I'm not sure how well the race conditions
> > will work out, but I assume that is a known issue.
> 
> There shouldn't be more race conditions on Woe32 than on Unix: On both kinds
> of systems, the signal handlers are run in the main thread, not in a separate
> thread.

I'm getting failures on mingw:

In file included from ../../lib/fatal-signal.c:30:
../../lib/sigprocmask.h:34: error: conflicting types for 'sigset_t'
/usr/lib/gcc/i686-pc-mingw32/3.4.4/../../../../i686-pc-
mingw32/include/sys/types.h:95: error: previous declaration of 'sigset_t' was 
here

The mingw <sys/types.h> system header contains this snippet:

#ifndef _SIGSET_T_
#define _SIGSET_T_
typedef int     _sigset_t;

#ifndef _NO_OLDNAMES
typedef _sigset_t       sigset_t;
#endif
#endif  /* Not _SIGSET_T_ */

It looks like you need to add a configure-time test whether sigset_t is 
available, before redeclaring it to unsigned int yourself.

-- 
Eric Blake