Mercurial > hg > octave-nkf > gnulib-hg
changeset 10236:d7f97416cbe1
Define SA_RESTART also for mingw.
author | Bruno Haible <bruno@clisp.org> |
---|---|
date | Sun, 22 Jun 2008 21:54:08 +0200 |
parents | a8938b62a368 |
children | 11dac0bf8a38 |
files | ChangeLog lib/sigaction.c lib/signal.in.h |
diffstat | 3 files changed, 13 insertions(+), 5 deletions(-) [+] |
line wrap: on
line diff
--- a/ChangeLog +++ b/ChangeLog @@ -1,5 +1,8 @@ 2008-06-22 Bruno Haible <bruno@clisp.org> + * lib/signal.in.h (SA_RESTART): New macro. + * lib/sigaction.c: Update comment. + * m4/sigaction.m4 (gl_SIGACTION): Require gl_SIGNAL_H_DEFAULTS. * m4/signalblocking.m4 (gl_PREREQ_SIG_HANDLER_H): New macro.
--- a/lib/sigaction.c +++ b/lib/sigaction.c @@ -35,11 +35,15 @@ the situation by reading static storage in a signal handler, which POSIX warns is not generically async-signal-safe. Oh well. - Additionally, SIGCHLD is not defined, so we don't implement - SA_NOCLDSTOP or SA_NOCLDWAIT; sigaltstack() is not present, so we - don't implement SA_ONSTACK; and siginterrupt() is not present, so - we don't implement SA_RESTART. Supporting SA_SIGINFO is impossible - to do portably. + Additionally: + - We don't implement SA_NOCLDSTOP or SA_NOCLDWAIT, because SIGCHLD + is not defined. + - We don't implement SA_ONSTACK, because sigaltstack() is not present. + - We ignore SA_RESTART, because blocking Win32 calls are not interrupted + anyway when an asynchronous signal occurs, and the MSVCRT runtime + never sets errno to EINTR. + - We don't implement SA_SIGINFO because it is impossible to do so + portably. POSIX states that an application should not mix signal() and sigaction(). We support the use of signal() within the gnulib