diff options
Diffstat (limited to 'libc/unistd/sleep.c')
| -rw-r--r-- | libc/unistd/sleep.c | 26 |
1 files changed, 21 insertions, 5 deletions
diff --git a/libc/unistd/sleep.c b/libc/unistd/sleep.c index 211c8434e..9db115a29 100644 --- a/libc/unistd/sleep.c +++ b/libc/unistd/sleep.c @@ -46,12 +46,28 @@ int main() { return 0; } * Testing on 2.4.20 and on 2.6.35-rc4: - * With SIG_DFL, nanosleep() is not interrupted by SIGCHLD. Ok. - * With dummy handler, nanosleep() is interrupted by SIGCHLD. Ok. - * With SIG_IGN, nanosleep() is NOT interrupted by SIGCHLD. + * With SIG_DFL, nanosleep is not interrupted by SIGCHLD. Ok. + * With dummy handler, nanosleep is interrupted by SIGCHLD. Ok. + * With SIG_IGN, nanosleep is NOT interrupted by SIGCHLD. * It looks like sleep's workaround for SIG_IGN is no longer needed? - * The only emails I can find are from 1998 - - * google for "sleep ignore sigchld". + * The only emails I can find are from 1998 (!): + * ---------- + * Subject: Re: sleep ignore sigchld + * From: Linus Torvalds <torvalds@transmeta.com> + * Date: Mon, 16 Nov 1998 11:02:15 -0800 (PST) + * + * On Mon, 16 Nov 1998, H. J. Lu wrote: + * > That is a kernel bug. SIGCHLD is a special one. Usually it cannot + * > be ignored. [snip...] + * + * No can do. + * + * "nanosleep()" is implemented in a bad way that makes it impossible to + * restart it cleanly. It was done that way because glibc wanted it that way, + * not because it's a good idea. [snip...] + * ---------- + * I assume that in the passed twelve+ years, nanosleep got fixed, + * but the hack in sleep to work around broken nanosleep was never removed. */ # if 0 |