Re: maint: Import texinfo.tex.

[Bruno Haible]

Hi Collin,

> I noticed that gperf distributes a very old texinfo.tex file from 2008.
> Doesn't it make more sense to import this from Gnulib where it is kept
> up-to-date?
> 
> Attached a proposed patch.

The patch takes always the newest, untested version of texinfo.tex and
uses it in the build process and for the release.

Which tests do we have for the result (gperf.pdf)? None!
  - No automated tests.
  - No manual tests (since the release manager will typically look only
    at the info and html output of the documentation, not at the PDF).

There are two lessons we ought to have learned from the Clownstrike disaster:

  - From the perspective of a software provider: Don't release untested
    stuff to all your customers!

  - From the perspective of a software consumer: Secure your supply chain!
    That is, don't blindly accept and integrate changes without even minimal
    testing on your side.

Since the only reasonable testing of texinfo.tex is manual — namely, take
the .pdf generated by the old and by the new texinfo.tex and compare them
visually —, your patch is not acceptable.

I took a texinfo.tex from the newest texinfo *release* (that's at least
better than a snapshot), did a comparison, and found that
  - there is an improvement: the text is darker and thus more readable,
  - there is a regression: the colour of interactive links was changed from
    brown to black. Giving a "smartphone UI" experience: you need to click
    (or at least hover) on a region in order to know whether it is interactive.
    The "desktop UI" experience, where the user can see where they may
    click, is definitely preferable.

So, I upgraded to the newest texinfo release and added a workaround for the
regression.
https://git.savannah.gnu.org/gitweb/?p=gperf.git;a=commitdiff;h=78a4fe8e9d9a89c0691a9e7d7163e50ab68e9542

Thanks for the report.

Bruno