On Wed, 28 Jan 2009 21:12:30 +0100 Michelle Konzack <linux4michelle@tamay-dogan.net> wrote: > Forgotten some things: > > 1) How do the RSS feed generator get the logfile? I asume that are > always the style like > http://www.emdebian.org/buildd/a/apt/trunk/apt-arm-1233120802.log > Should this be a command line tool or a webscript? http://www.emdebian.org/buildd/$p/$package/trunk/$package-$arch-$timestamp. Now that you mention it, it is the timestamp that causes difficulty with retrieval. It will be easy once the generator is on the same server. I can look into a webscript that just lists the paths to the most recent build logs for each package. I'll need to make a script for the Grip build logs too. http://www.emdebian.org/grip/logs/ I've got a simple PHP script that can render a series of links like this: print "<a href=\"logs/$file\">$timestr</a><br/>"; <a href="logs/1233057601.log">Tue 27th Jan, 12:00:01 GMT</a><br/> Would a different identifier be more suitable? Would a simple path without the link be more useful? > > or auto-process to say: > > $count packages updated in Crush > > 2) $count is what? > Do you count new uploads in a timeframe like one week? The count of build logs uploaded on the same day - the autobuilder runs only once per 24 hours and uploads build logs once each build is complete. The script mentioned above could isolate the logs according to the day of the upload by converting the timestamp, making the $count trivial. > I prefer to build small *.in files like > 1234567890_$pkg.in > where 1234567890 is the buildtime and then if I prebuild the > RSS-feed I compare the time, whether it is in the timeframe. OK. > > The repeated for each package uploaded since the last run of the feed > > generator script. > > Hm, I was ongoing to build the RSS feed immediately... because each time > a new buildlog comes in I generate a timestamed IN file which contain > the stuff between <item></item> and then I 'cat' it together which takes > on a decent server under 1 second even for 100 packages since they have > only some bytes. > > Q: How many builds do you have per day? One build sequence per architecture, unknown number of builds per sequence. How often do you poll the server to determine whether a new buildlog has arrived? This could be easier once the generator is on the same server. > If not build immediately, how often do you want to build the RSS feed? Once per 24 hours? Maybe increase that to every few hours once there are more architectures supported by Crush. Crush uploads are simple to divide into individual items, Grip is harder. -- Neil Williams ============= http://www.data-freedom.org/ http://www.linux.codehelp.co.uk/ http://e-mail.is-not-s.ms/
Attachment:
pgpE44Mvn47Xu.pgp
Description: PGP signature