From: Tomi Ollila <tomi.ollila@iki.fi>
To: "W. Trevor King" <wking@tremily.us>, David Bremner <david@tethera.net>
Subject: Re: [PATCH] doc: Allow rst2man.py as an alternative to rst2man
In-Reply-To: <20150103164506.GR14967@odin.tremily.us>
References: <1420294479-27224-1-git-send-email-tomi.ollila@iki.fi>
	<1420300167-5060-1-git-send-email-david@tethera.net>
	<20150103164506.GR14967@odin.tremily.us>
User-Agent: Notmuch/0.19+14~ge04617b (http://notmuchmail.org) Emacs/24.3.1
	(x86_64-unknown-linux-gnu)
Date: Sat, 03 Jan 2015 19:12:13 +0200
Message-ID: <m2oaqfbxg2.fsf@guru.guru-group.fi>
MIME-Version: 1.0
Content-Type: text/plain
Cc: notmuch@notmuchmail.org
Precedence: list

On Sat, Jan 03 2015, "W. Trevor King" <wking@tremily.us> wrote:

> On Sat, Jan 03, 2015 at 04:49:27PM +0100, David Bremner wrote:
>> At the risk of bike-shedding, rewrite the configure check to be more
>> "obvious". This is pretty subjective, but in particular I'm not a
>> big fan of resetting RST2MAN at the bottom of every iteration of the
>> for loop.  Also conform a bit more to the convention of using a
>> lower case variable in configure and assigning it in Makefile.
>
> I'm fine with all of these changes.

I am also -- and I'd prefer lower-case variables -- however the convention
of using lower case variables is not so clear here (or weren't until
David's ${python} path ;) -- we have mixup of upper & lower case variables
there, and otoh, other (auto)configure scripts seems to use *_RST2MAN.

But, instead of working this further we could also consider Jani's irc comment
of dropping rs2man support altogether... or just keep the status quo and
instruct users just install (recent enough) sphinx to their systems to get
man pages created...

Tomi

PS: In one Scientific Linux 6.2 system I build and run notmuch I have

  # Whether there's a sphinx-build binary available for building documentation
  HAVE_SPHINX=0

  # Whether there's a rst2man binary available for building documentation
  HAVE_RST2MAN=1

because sphinx-build is not recent enough -- sphinx-1.0-build would be but
as it cannot be used by default and as rst2man (w/o .py) works I have not bothered.
I'd presume anyone else using such an old system will have bigger problems
getting their notmuch compiled (recent enough zlib being one of those), so
this makes me wonder whether there is actually anyone needing rst2man there
(besides me, of course :)