I'm flattered Autrijus, thanks.
What I'm doing using my .forward file to pipe into a perl script (mail_filter.pl). The info is then put into a PostgreSQL db (dbSchema.sql), which is then accessed from a cron job (cron_smoke.pl) which picks out modules at least 1 hour old (from the PAUSE sent date), and smokes them. dbSmoke.pm is just a little module which contains some common db routines shared between cron_smoke.pl and mail_filter.pl. I came up with this as a solution to the forking and sleeping for 1 hour then testing. On my machine, this was causing Postfix to believe the commands were timing out, and everything was being bounced back, but processed correctly locally. In addition, using the db effectively serializes the smoking, which forking could not do. I run the cron job every hour, so as long as I can smoke new modules in under an hour, I'm fine. Deamonizing the cron process to sleep based on the next modules testing time would be a good solution, but I'm lazy, and it's working now. It's a 2 hour job, so it's not documented, nor clean, and is lacking features, but it works. One day, I'm going to export the testing reports from the db to my local Apache server so I can monitor the status of testing without looking through mail logs (which I don't do). The mail account doing the testing is dedicated to this, so I don't worry about anything else being sent there Code is at: http://www.logicalchaos.org/perl/smokin.tgz Feel free to improve, etc... but please send improvements back to me so I can use them :-) > On Mon, May 13, 2002 at 08:06:43PM +0200, Jost Krieger wrote: > > I'm trying to set up a smoking server and get a few funny results: > > 2. Now that I got a few failures it turns out that cpantest > > doesn't pick up the error messages (see cpan-testers). > > My (unsupported) idea is that it might be a result of my running: > > xargs cpansmoke -adpsv >/tmp/smokeout6 2>&1 < /tmp/smoking & > > Well... I think you can ask Robert Creager for his Automated > configuration. (CC'ed to him.) > > > with /tmp/smoking filled from cpan-testers (currently manually, this > > will change). > > I'm not sure what arguments will xargs take exactly... If you are > feeding it a mailbox, I think dangerous things would happy as each > line in the subject body will be taken as a module to test. > > Before I lost my production server (ugh), I was using the mail > filtering method I described in the CPANPLUS::TesterGuide, so I have > no idea about other possible arrangements. I hope to resume smoking > on freebsd once our HD gets restored in a few days. > > > I fear that smoking many modules at once might be a problem. > > Hmm. cpansmoke handles the arguments sequencially, so it's not a > problem. Even if you're forking, my experience is that the speed of > CPAN upload has not reached to point that can tax my machine too much. > (YMMV). > > Hope that helps. > > Thanks! > /Autrijus/ >
