From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 22 Aug 2012 18:24:50 GMT
Subject: Re: Porting git to HP NonStop
Message-ID: <7v393e943x.fsf@alter.siamese.dyndns.org>
In-Reply-To: <003d01cd8090$214ca710$63e5f530$@schmitz-digital.de>

"Joachim Schmitz" <jojo@schmitz-digital.de> writes:

>> Nice.  And we have xmemdupz() would be even better as you followed-up.
>
> How's that one used?

I forgot that we frown upon use of any x<allocate>() wrapper in the
compat/ layer as J6t mentioned.

So probably something along these lines...

	int retval;
	char *dir_to_free = NULL;
	size_t len = strlen(dir);

        if (len && dir[len - 1] == '/') {
		dir_to_free = malloc(len);
                if (!dir_to_free) {
			fprintf(stderr, "malloc failed!\n");
			exit(1);
		}
                memcpy(dir_to_free, dir, len - 1);
                dir_to_free[len - 1] = '\0';
                dir = dir_to_free;
	}                
	retval = mkdir(dir, mode);
	free(dir_to_free);
        return retval;

It might be possible to for the error path to get away with
something like:

	if (!dir_to_free)
		return -1;

if we know the callers are prepared to see mkdir() failing with
ENOMEM, but that is not very likely.