This document describes the proper operation of the glxinfo
repository. The result can be viewed at
https://people.freedesktop.org/~imirkin/glxinfo/

The idea is that each hardware/version combination has a separate
file. This file is loaded on demand by the main glxinfo.js logic,
which knows which file each hardware/version combination lives
in. This is in the VERSIONS map towards the top. Make sure to add a
proper entry to VERSIONS -> Lib name (e.g. Mesa 12.0.0) -> hw name.

In order to generate a javascript-ized version of a glxinfo output,
you can use the parse-glxinfo.py tool. This will either run glxinfo
locally, or, if an argument is supplied, parse the given file. This
file should contain the output of 'glxinfo -l -s'.

Unfortunately this tool is not perfect, and a few fixups have to be
applied manually:

 - For radeon, change the first argument of the register() call to "radeon"
 - Manually adjust the library version to match what is specified in glxinfo.js
 - Manually adjust the ES gl/glsl/library versions, as the parser
   doesn't handle them very well
 - Rewrite any large values in M or G - mostly the
   MAX_TEXTURE_BUFFER_SIZE, and a few odd 2G-sized values reported by
   softpipe.

It's a good idea to diff the new file against an older version to
ensure that no such small fixups have been missed.

In order to test your changes, run

  python -m SimpleHTTPServer

or

  python3 -m http.server

in the directory containing index.html, and check
http://localhost:8000/ in your browser to make sure that everything
displays reasonably.

After pushing your change, you can go into
~imirkin/public_html/glxinfo on annarchy and do a "git pull". This
should update the live site.