2009/4/30 Johannes Schindelin <Johannes.Schindelin@xxxxxx>:
> Hi,
>
> On Wed, 29 Apr 2009, Alex Riesen wrote:
>
>> 2009/4/29 Junio C Hamano <gitster@xxxxxxxxx>:
>> > +
>> > + if (S_ISGITLINK(mode)) {
>> > + blob = xmalloc(100);
>> > + *size = snprintf(blob, 100,
>> > + "Subproject commit %s\n", sha1_to_hex(sha1));
>>
>> snprintf returns a signed value. It also has a bad record of returning
>> negative values for obscure reasons (on obscure platforms, admittedly).
>>
>> For this particular case,
>>
>> strcpy(blob, "Subproject commit ");
>> strcat(blob, sha1_to_hex(sha1));
>> strcat(blob, "\n");
>> *size = strlen(blob); /* that's a constant */
>>
>> could be considered.
>
> Actually, we know _exactly_ the size of the thing. It is 18+40+1. But I
> think that *size wants to have the size, not the length. So add 1.
>
> In any case, I don't think that we have to jump through hoops here:
> snprintf() is _most_ unlikely to return something negative here. So I'd
> say that readability trumps paranoia here.
>
http://www.google.com/search?q=snprintf+negative+return+value
First link: http://bytes.com/groups/c/590845-snprintf-return-value
Look for "(Windows, mingw)"
--
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