On Thu, Apr 05, 2012 at 05:34:36PM -0500, Jonathan Nieder wrote:
> > --- a/run-command.c
> > +++ b/run-command.c
> > @@ -117,8 +117,11 @@ static int exists_in_PATH(const char *file)
> >
> > int sane_execvp(const char *file, char * const argv[])
> > {
> > - if (!execvp(file, argv))
> > - return 0;
> > + /*
> > + * No need to check the return value; if it returns at all, an error
> > + * occurred.
> > + */
> > + execvp(file, argv);
> >
>
> I'd rather have the "return 0" then. Such a comment that focuses on C
> library details rather than providing additional information for
> understanding the sane_execvp function is distracting.
I think both (attempt to) document the same thing: that if we get past
execvp, we know are in error-checking mode. Which is not explicitly said
anywhere. So maybe:
diff --git a/run-command.c b/run-command.c
index 7123436..e6ece79 100644
--- a/run-command.c
+++ b/run-command.c
@@ -117,10 +117,12 @@ static int exists_in_PATH(const char *file)
int sane_execvp(const char *file, char * const argv[])
{
- if (!execvp(file, argv))
- return 0;
+ execvp(file, argv);
/*
+ * If we are still running, we know an error occurred; let's try to
+ * diagnose it more specifically.
+ *
* When a command can't be found because one of the directories
* listed in $PATH is unsearchable, execvp reports EACCES, but
* careful usability testing (read: analysis of occasional bug
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html