From: Junio C Hamano <junkio@cox.net>
Date: Fri, 07 Apr 2006 20:35:30 GMT
Subject: Re: realloc
Message-ID: <7vmzexi0od.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <118833cc0604071311v1da93f83n112cc2ea44552ca9@mail.gmail.com>

"Morten Welinder" <mwelinder@gmail.com> writes:

> I could be wrong, but shouldn't
>
>       var = realloc (var, whatever);
>
> be changed to call xrealloc?  That, or assign to a different variable and check
> for NULL.
>
> This should affect the last four hits below.
>
> M.
>
>
>
>
> /scratch/welinder/git> grep -w realloc *.c
> daemon.c:               newlist = realloc(socklist, sizeof(int) *
> (socknum + 1));
> diff-delta.c:                           out = realloc(out, outsize);
> git.c:          cmdname = realloc(cmdname, cmdname_alloc * sizeof(*cmdname));
> ls-files.c:             which->excludes = realloc(which->excludes,
> sha1_file.c:                            buf = realloc(buf, size);

There is no excuse for not using xrealloc() in git.c,
ls-files.c, and sha1_file.c.

The diff-delta.c code wants to be independent from the rest of
git code, so it probably should check the returned value itself.

Historically to a certain degree daemon.c also wanted to be
independent from the rest of git, but I suspect it still is the
case (it uses small pieces from packet interface but that is
about it).