On Mon, Feb 17, 2025 at 6:30 PM Evan Gates <evan.ga...@gmail.com> wrote:
> On Fri, 14 Feb 2025 at 09:25 Tavian Barnes, <taviana...@tavianator.com> wrote:
>
> > ---
> > find.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/find.c b/find.c
> > index ce551e5..71488da 100644
> > --- a/find.c
> > +++ b/find.c
> > @@ -244,6 +244,7 @@ spawn(char *argv[])
> > switch((pid = fork())) {
> > case -1:
> > eprintf("fork:");
> > + return -1;
>
> eprintf calls exit(1), there's no return to do here
Ah okay, makes sense.
> > case 0:
> > execvp(*argv, argv);
> > weprintf("exec %s failed:", *argv);
> > @@ -252,7 +253,7 @@ spawn(char *argv[])
> >
> > /* FIXME: proper course of action for waitpid() on EINTR? */
> > waitpid(pid, &status, 0);
> > - return status;
> > + return WIFEXITED(status) && WEXITSTATUS(status) == 0 ? 0 : -1;
>
> From POSIX wait() documention:
>
> The value stored at the location pointed to by stat_loc shall
> be 0 if and only if the status returned is from a terminated
> child process that terminated by one of the following means:
>
> 1. The process returned 0 from main().
>
> 2. The process called _exit() or exit() with a status
> argument of 0.
>
> 3. The process was terminated because the last thread
> in the process terminated.
>
> Why do extra work if we already have 0 status on success?
I didn't realize POSIX specified the encoding of a successful exit,
neat. On further testing I can see this patch doesn't change
anything. I'm not sure why I thought it was necessary in the first
place. Please drop it.
> > }
> >
> > static int
> > --
> > 2.48.1
> >
> >
>
