From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Sat, 10 Nov 2007 13:47:36 GMT
Subject: Re: [PATCH] builtin-commit: fix --signoff
Message-ID: <Pine.LNX.4.64.0711101346120.4362@racer.site>
In-Reply-To: <7vr6iyo4ff.fsf@gitster.siamese.dyndns.org>

Hi,

On Sat, 10 Nov 2007, Junio C Hamano wrote:

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> >  	if (signoff) {
> > ...
> > +		strbuf_init(&sob, 0);
> > +		strbuf_addstr(&sob, sign_off_header);
> > +		strbuf_addstr(&sob, git_committer_info(1));
> > +		p = strrchr(sob.buf, '>');
> > +		if (p)
> > +			strbuf_setlen(&sob, p + 1 - sob.buf);
> > +		strbuf_addch(&sob, '\n');
> > +
> > +		for (i = sb.len - 1; i > 0 && sb.buf[i - 1] != '\n'; i--)
> > +			; /* do nothing */
> > +		if (prefixcmp(sb.buf + i, sob.buf))
> > +			strbuf_addbuf(&sb, &sob);
> >  	}
> 
> At this point doesn't this leak sob.buf?

Darn.  I had a "strbuf_release(&sob);" there, but at some stage removed it 
by mistake.

But there is more to be fixed.  Rather than let git_committer_info() add 
the timestamp, only to strip it away, is a waste.

Will redo the patch.

Ciao,
Dscho