Re: [PATCH v2 0/3] docs, parallelism: Rearrange how jobserver reservations are made

From: Jonathan Corbet <corbet@lwn.net>
To: Kees Cook <keescook@chromium.org>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>,
	linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 0/3] docs, parallelism: Rearrange how jobserver reservations are made
Date: Fri, 22 Nov 2019 10:39:32 -0700	[thread overview]
Message-ID: <20191122103932.65d79f1f@lwn.net> (raw)
In-Reply-To: <20191121205929.40371-1-keescook@chromium.org>

On Thu, 21 Nov 2019 12:59:26 -0800
Kees Cook <keescook@chromium.org> wrote:

> v2:
>     - correct comments and commit logs (rasmus)
>     - handle non-parallel mode more cleanly (rasmus)
>     - reserve slots 8 at a time (rasmus)
> v1: https://lore.kernel.org/lkml/20191121000304.48829-1-keescook@chromium.org
> 
> Hi,
> 
> As Rasmus noted[1], there were some deficiencies in how the Make jobserver
> vs sphinx parallelism logic was handled. This series attempts to address
> all those problems by building a set of wrappers and fixing some of the
> internal logic.
> 
> Thank you Rasmus for the suggestions (and the "jobhog" example)! :)

OK, I have applied this set for 5.5.

I do worry that this all looks a little complex and fragile.  I wonder if
there's a way that we could set up some sort of dependency chain that
would just tell make not to run the docs builds in parallel with anything
else?  That is more-or-less the effect of what we're doing anyway.

Meanwhile, though, this seems to work, so let's go with it :)

Thanks,

jon