On 29/09/2009 19:35, Yaakov (Cygwin/X) wrote:
Anyway, to answer the question, AFAICS in glibc, <signal.h> #include
<bits/types.h> unconditionally[1]. (<sys/signal.h> is just one line:
#include <signal.h> [2])
So should I take the first route, patching newlib instead?
OTOH, this comment in the offending hunk of <sys/signal.h> (which is
only for Cygwin and RTEMS) makes me wonder:
/* The first argument to kill should be pid_t. Right now
<sys/types.h> always defines pid_t to be int. If that ever
changes, then we will need to do something else, perhaps along the
lines of <machine/types.h>. */
int _EXFUN(kill, (int, int));
int _EXFUN(killpg, (pid_t, int));
Is that supposed to mean that we don't want to use pid_t here at all,
and the intended solution would be to change killpg to (int, int), as
ugly as that is, leaving only <cygwin/signal.h> needing the #include
<sys/types.h>?
Yaakov
