[PATCH 3 of 6] formatter: load templates section like a map file

[Yuya Nishihara]

On Fri, 16 Jun 2017 10:59:28 -0700, Martin von Zweigbergk wrote:
> On Fri, Jun 16, 2017 at 9:37 AM, Yuya Nishihara <yuya at tcha.org> wrote:
> > # HG changeset patch
> > # User Yuya Nishihara <yuya at tcha.org>
> > # Date 1492859695 -32400
> > #      Sat Apr 22 20:14:55 2017 +0900
> > # Node ID 32ea66830e68419c6c97c0bcfc53f053f628b399
> > # Parent  06b1ac59547e8ef086bae9bc51740318c7661e7e
> > formatter: load templates section like a map file

> > -  $ echo '[templates]' >> .hg/hgrc
> > -  $ printf 'simple = "{rev}\\n"\n' >> .hg/hgrc
> > -  $ printf 'simple2 = {rev}\\n\n' >> .hg/hgrc
> > +  $ cat <<'EOF' >> .hg/hgrc
> > +  > [templates]
> > +  > simple = "{rev}\n"
> > +  > simple2 = {rev}\n
> > +  > rev = "should not precede {rev} keyword\n"
> 
> Why would you want to do this? In other words, why don't we
> reject/warn/ignore that instead? I'm guessing it's because we already
> have supported this form, so we don't want to break BC?

templates.rev, the definition, is still valid since it can be referenced
by '-T rev'. As the built-in keywords space may grow, we can't reliably
reject template definitions conflicting with the future builtins.

Another reason is that hgweb templates rely on this weird behavior.