Hi, On 23/02/12 19:43, Diego Biurrun wrote: > On Mon, Feb 13, 2012 at 09:32:04AM +0200, Miika Komu wrote: >> On 02/13/2012 09:14 AM, Diego Biurrun wrote: >>> On Mon, Feb 13, 2012 at 07:47:31AM +0200, Miika Komu wrote: >>>> Btw, hipl_autobuild.sh is still semi-automated and a bit tedious to set >>>> up. It is designed to be run from crontab rather than effortless >>>> execution from the command line. >>> >>> No, no, no, it was never "designed" to be run from cron. The cronjob >>> that runs the autobuilder does run from cron, true, but all it does is >>> call tools/hipl_autobuild.sh from the command line. >> >> Why is the emphasis then on the cron execution? Observe the >> beginning of tools/hipl_autobuild.sh: >> >> # HIPL autobuild script for periodic compilation and quality tests. >> .. >> # The script is suitable to be run from cron in order to provide basic > > I'll rephrase that to put less emphasis on the cronjob part. Still my > point remains valid: I did not intend for this to be run from cron only. > I tested it from the command line while creating it and I often use it > from the command line to test my branches. > >>> Previously it would expect a command line parameter to indicate the branch >>> it was supposed to test, but I recently changed it to derive a suitable >>> default from Bazaar. It also used to expect a certain home directory >>> layout in order to run the OpenWrt and MAEMO cross-compilation tests. >>> Now it bails out with a message if those are not found, but even before >>> it would run all the previous tests just fine. >> >> Thanks for the history lesson but this doesn't still work out of the box: >> >> mkomu@bling:~/projects/hipl-bzr/trunk$ tools/hipl_autobuild.sh >> usage: tools/hipl_autobuild.sh <branch_name> >> mkomu@bling:~/projects/hipl-bzr/trunk$ tools/hipl_autobuild.sh trunk >> bzr: ERROR: Not a branch: "/home/mkomu/src/hipl/trunk/". >> cat: /home/mkomu/tmp/autobuild/hipl/HIPL_REVISION_trunk: No such >> file or directory >> >> Feel free to say "fix it yourself" or "you're too lazy/stupid/busy >> to figure to this out" but don't give any excuses why this is >> perfect as it is because it isn't. > > Should all be fixed. Thanks for the criticism, let me know what you > think of the current state. the current state is awesome! Thanks. >>>> IMHO, the best way to improve would be to integrate this somehow into >>>> "make alltests". A better than nothing alternative is to document >>>> copy-paste-instructions to HACKING explaining "what do I need to >>>> execute from the command line" to get it working instead forcing all >>>> developers to read the source code of the script and set it up using >>>> the trial-and-error method. >>> >>> The script has a verbose documentation block at the top. Duplicating >>> that I do not consider a good idea, it will only get out of sync. >> >> IMHO, the script documents the design but not the usage. >> >> Based on this comment, we don't even need the HACKING or manual >> because all you need is the source code. > > Documentation should not be duplicated. It is hard enough to keep it > up to date in one place. If it is available in two places, I bet > people will alternate updating either, so both will be broken. > > Please let me know what extra documentation you want and where you > want to see it. It should now work out of the box and not need much > documentation. Nothing else to complain than the missing chmod a+x tools/hipl_autobuild.sh :)