Hi, On 02/13/2012 09:14 AM, Diego Biurrun wrote:
On Mon, Feb 13, 2012 at 07:47:31AM +0200, Miika Komu wrote:On 13/02/12 00:26, Diego Biurrun wrote:True, but you should also not rearrange things w/o need to do so. I'll try to update HACKING this week, there's a bunch of stuff I wanted to add in there. ....I didn't know there are other testing methods like 'make alltests' and autobuilder, but before every commit I make sure the code pass the 'make check'. from now on I will also check 'make alltests'. Could you give some hints about how to use the autobuilder you mentioned?Who is mentoring you over in Finland and why are they not giving you appropriate hints? :) The autobuilder is the script tools/hipl_autobuild.sh, it should work if you run it directly. You could also just read Makefile.am to find out about available targets. You could also read the output of configure to see which options are enabled or not. Also look at 'configure --help' for available options.actually, I mistakenly assumed that alltests and hipl_autobuild.sh were already... were already .. what? Sorry, I honestly don't know what you are trying to say here ...
...already documented in HACKING. (Sorry, I swallowed the end :)
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
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.
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.