From mboxrd@z Thu Jan  1 00:00:00 1970
From: john at keeping.me.uk (John Keeping)
Date: Sun, 7 Apr 2013 13:36:22 +0100
Subject: [PATCH 14/19] ui-summary.c: use struct strbuf instead of
 fixed-size buffers
In-Reply-To: <CAHmME9ozyyb1Y=_e=NAGtvX-JDXyMEMHakEh-6XhTGYm1zbopw@mail.gmail.com>
References: <cover.1365326321.git.john@keeping.me.uk>
 <4dd4fc55af2528bc676893c2bff8163d7f7d21d5.1365326321.git.john@keeping.me.uk>
 <CAHmME9ozyyb1Y=_e=NAGtvX-JDXyMEMHakEh-6XhTGYm1zbopw@mail.gmail.com>
Message-ID: <20130407123622.GY2222@serenity.lan>

On Sun, Apr 07, 2013 at 02:20:26PM +0200, Jason A. Donenfeld wrote:
> On Sun, Apr 7, 2013 at 11:30 AM, John Keeping <john at keeping.me.uk> wrote:
> >         /* Prepend repo path to relative readme path unless tracked. */
> > -       if (!ref && *ctx.repo->readme != '/')
> > -               ctx.repo->readme = xstrdup(fmt("%s/%s", ctx.repo->path,
> > -                                              ctx.repo->readme));
> > +       if (!ref && *ctx.repo->readme != '/') {
> > +               struct strbuf buf = STRBUF_INIT;
> > +               strbuf_addf(&buf, "%s/%s", ctx.repo->path, ctx.repo->readme);
> > +               ctx.repo->readme = strbuf_detach(&buf, NULL);
> > +       }
> 
> I do in fact see the merits of using strbuf and the purpose of this
> patch set, but OTOH, this little chunk here illustrates the downsides
> -- the extra verbosity this adds is really a bummer. Perhaps this
> could be moved into a new fmtalloc helper function, if this pattern is
> common?

That could be sensible.  I didn't do that initially because a lot of the
uses are just:

    allocate
    format
    use
    free

and there is no real saving from a helper in that case, but in case
where we format an then detach the helper could be useful.  I'll do that
in the reroll.