Not too much has happened this month, at least not to rave about. A lot of backend fixes and abstraction of common modules into their own distribution, more tests and generally a bit of a clean up. Following some runs with Devel-NYTProf, it highlighted a few areas where the Reports website generation could be sped up. At first it didn't appear to make much of a difference, but after a week trying to catch up with itself, the generation code did start to improve considerably. To the point we now have as many as 4 updates of the backend data and websites each day. From once taking as long as 18 hours to do a daily update, the updates now take between 3-6 hours for each run. I may look at making this a little more frequent, but for now I would like to keep it stable so I can see how we fare with the current level of reporting.
The pressure was off somewhat last month, so it was nice to have a stable base to watch and see if anything wasn't working as it should. It did highlight a couple of minor glitches, including incorrect permissions on the OpenThought and CGI files, but thankfully no major problems. However, that isn't to say there hasn't been a major problem, as there has, you just likely never noticed. Unfortunately the rsync to mirror CPAN, which updates the CPAN Testers BACKPAN mirror and allows CPAN Testers to run the analysis for the Uploads database, failed to update for over a week. It appears that too many people are trying to rsync from the master CPAN mirror at FUNET, and some requests were not able to be actioned. It has highlighted that the mirroring infrastructure perhaps does need a bit of a rethink. Some quite likely could be quite comfortable accessing a second tier mirror perhaps once a day, and those of us who would like to be able to update more frequently can act as secondary tiers. At the moment only BACKPAN is available from the cpantesters.org domain, but I shall be looking to allow it to also provide a public CPAN mirror too. Plus with the ability to update the data 4 times a day, I'd like to have the mirrors update 4 times a day too. However, until the guys at FUNET are comfortable with that, it'll stick to once a day for now.
I have started work on the dynamic Reports website, and it has already thrown up a bit of a quandry. How do I detect whether a request is for a Javascript enabled browser or not? It seems there is no easy method, as even though a browser may be able to handle javascript, the user may have switched it off and nothing in the HTTP request indicates that. The suggestion in many places seems to be set a cookie via javascript, but that only works if you have some sort of landing page that refreshes before loading the real page. Personally I don't like this, so I'm looking to changing the URLs that will be used. The current URLs will map to the Javascript versions, with the flat HTML pages being available via a slightly different link. If there is a better way, I'd love to hear about it.
As I've been mentioning for a while now, the 2009 QA Hackathon event will be taking place in Birmingham, UK from Saturday March 28 to Monday March 30. 3 full days of hacking on TAP, CPAN Testers and other QA and testing projects are planned, with currently 20 people in attendance. Several key collaborators involved with CPAN Testers will be there, so we're hoping that we can finally get CT2.0 up and working for everyone as soon as possible after the event.
We topped 143 testers submitting reports last month, so thank you once again to everyone involved. The mappings this month included 27 total addresses mapped, of which 12 were for newly identified testers.
