Re: Testing lots of Texinfo files

[Gavin Smith]

On Thu, Jan 21, 2021 at 07:04:19PM -0600, Jacob Bachmeyer wrote:
> Gavin Smith wrote:
> > There are many manuals that I couldn't get easily as there wasn't a
> > link to download the Texinfo sources on the website (where there was,
> > this was always because the web manuals were generated with the
> > gendocs script). It would be a never-ending project to try to get
> > everything and the dependencies for all the manuals, but if there are
> > other manuals that are important to check that aren't listed below,
> > you could let me know if you know how to get hold of them.
> 
> As a maintainer of DejaGnu, the DejaGnu manual is important to me and
> getting an up-to-date copy just might point the way to getting more Texinfo
> sources.  The latest revision is available at
> <URL:https://git.savannah.gnu.org/cgit/dejagnu.git/tree/doc/dejagnu.texi>
> and its only dependencies are "version.texi" (which is common with
> Automake-using projects and can be easily stubbed out) and "fdl.texi" (the
> text of the GFDL).
> 
> I expect that many GNU projects are likely to have manual sources in
> analogous locations on Savannah.

Thanks, I added your manual.  (It raised no issues.  Only odd thing
in the manual was that you had "@documentencoding us-ascii", but it
was all ASCII anyway.  The Texinfo manual documents that this has
no effect, so it's not clear to me whether the encoding should be
treated as UTF-8 or as ISO-8859-1 (as both include ASCII as a subset).)

I wasn't sure if it would be the case, but all the git repositories
are listed at:

https://git.savannah.gnu.org/cgit/

I guess this is an interface to git repositories so the files you see
there may not be easily downloadable with a script without actually
cloning the git repository themselves, which would use a lot of
network data.  You could probably write a script to do a shallow git
clone of some of these repositories, especially where they have a
doc subdirectory.  (Over 1000 are listed there). Preferrably in a way
that doesn't use 10's or even 100's of GB of network traffic.

It does seem like a sensible thing to do, rather than inadvertently
breaking some manuals and waiting for users to complain about it:
if issues can be discovered before a release, why not do so?

Most manuals will be fairly similar in the functionality that they
exercise, but "quantity has a quality of its own" and processing a
large amount of manuals is likely to bring things up.