Cross-posted from the CPAN Testers Blog.
July has seen several major changes to the backend code to running the CPAN Testers Reports website. The builder got some rules added to help manage the updates better and the front page now gets a status table. Tim Bunce suggested having some form of status indicator, and at the time some of the data needed to evaluate any indicators wasn't available. However, with some changes to the database, I was able to get the right information. As the site takes up to 5 days to get through all the changes, I had to wait a week before making the status table visible. At the moment the builder seems to have settled on running 2-3 days behind at most, and high profile pages often get built within a few minutes. The database now gets updated hourly, so there is now a much faster turn around of some of the pages. In addition to the status table, every author and distribution page now includes a timestamp of when the page was last built. If it's waiting in the queue to be rebuilt, at least you can now see how old the data available is. In many cases it will only be a few hours.
There are currently 25k+ pages on the site (double if you include the static site), with roughly 6 files being generated for each request. This is now streamlined as much as possoble, but I'm still looking at better ways to improve the process of creating them. With so much data now available it'll become more important to figure that one out.
Another change has been to the CPAN Testers Statistics website. For the last year or so, I've been asked various questions about the data, and how different sets of criteria stack up. Some I already provide regarding the CPAN Testers, but there is also a lot of data regarding CPAN too. As such, there are now some extra pages included on the Statistics website, that give a current snapshop of the state of CPAN. I've been meaning to provide these CPAN graphs for some time, and was a suggestion given to me after a London Perl Workshop, in 2007! If there are any other trend graphs or stats tables that you think might be useful or interesting, please let me know. I have another couple of tables I'm planning to add, but I'm always open to adding more :)
This week I'll be at YAPC::Europe 2009 in Lisbon. If you're a CPAN Tester and will be in Lisbon too, please come and say hello. I was disappointed to not meet a couple of people in Pittsburgh, though I did get to meet our 4 millionth poster to the cpan-testers mailing, George Greer. So please don't be shy, just come and find me and introduce yourself :)
Last month we had a total of 171 testers submitting reports. The mappings this month included 22 total addresses mapped, of which 10 were for newly identified testers.
Much the same as last month, though congrats to Dan Collins, who has been outstanding in submitting reports. There was a time I thought no-one could compete with Chris, but it seems he does indeed have a competitor. However, for anyone thinking that we have all scenarios covered, please take a look at the monthly test coverage. We are still looking for a variety of platforms, and indeed some of the older perls too. Unfortunately we don't know what is out there in the wild, and concentrating on the latest and greatest is not really a true representation of what companies are actually still using. So if you want to get involved with CPAN Testers, and don't have an unusual platform to test on (if you do, *please* get involved), you can still help out by ensuring that new CPAN releases still work on older version of perl. You might be surprised to learn that deployment is still happening on 5.003! While the author might not be willing to support older versions, many users still need that information.
We're fast approaching the 5 millionth post to CPAN Testers, and expect to see it hit sometime during August month. With so many reports being posted at the moment, I'm also expecting the 5 millionth report to also hit slightly later in August. We shall see by next month's summary :)
I tried some time ago testing on 5.5, I was able to install CPAN::Reporter that time (now you won't be able to install fresh CPAN.pm there), but somehow Safe.pm got upgraded and it stopped working.
Later I tried 5.4, it was hard even to install it, and upgrading CPAN.pm is very hard, so I was not able to install even CPAN::Reporter.
I guess it up to real users of these versions to try to make reporting toolchain working. They deploy their software, but do they install CPAN modules?
Re:older perls
barbie on 2009-08-04T13:37:51
but do they install CPAN modules?
From personal experience, yes. However, in some instances BACKPAN is used to get the tarball of a older known working version
:)
