Make watching GNOME bugs more efficient by doing just one buglist.cgi request

[Olav Vitters]

Malone currently watches Bugzilla bugs by using buglist.cgi very inefficiently. It makes a request for every bug. This takes 1 - 10 seconds _per_ bug. For all bugs this takes about 6 minutes total. Malone says there it watches 101 GNOME bugs, so 3.5 sec average per bug.

Malone should just do one buglist.cgi request.

Currently it requests:
http://bugzilla.gnome.org/buglist.cgi?bug_id=321366&form_name=buglist.cgi&ctype
=rdf&bug_id_type=include

It should combine the bug numbers with ",", like:
http://bugzilla.gnome.org/buglist.cgi?bug_id=321366,163163&form_name=buglist.cgi&ctype
=rdf&bug_id_type=include

Bugzilla allows you to use POST (handy when you have a lot of bug numbers).

[Christian Reis (kiko)]

Fixing this will be interesting -- the current externalsystem.py code
is built with the assumption that it will handle one bug at a time --
get_bug_status accepts a single bug_id as an argument, and that's what
issues the buglist.cgi request. To fix this we need to factor out most
of the code in checkwatches.py and move it all into a single
ExternalSystem.update_watch_status() call -- instantiating one per
bugtracker we want to check.

[James Henstridge (jamesh)]

It'd probably be worth looking at reworking the status update code to be multithreaded: sort the bug watches to be updated into queues (one per bugtracker), and have a thread for each queue.

This would be faster for us too, without adding more load on the remote bug trackers.

[Stuart Bishop (stub)]

I've disabled the checkwatches.py script on production on recommendation by seb128. It isn't nice to DOS the Gnome bugzilla every day.