Re: More smoke questions
[email protected] (Alan Burlison) Thu, 07 Aug 2003 15:29:09 +0100
| Newsgroups | perl.daily-build |
|---|---|
| Message-ID | <[email protected]> |
Abe Timmerman wrote: > Sure, that is probably the only way to keep your local copy of the source-tree > up to date. You only need to take care that you don't start a smoke while > doing that. That will work fine - thanks. >>2. How do I set flags that I would normally pass to Configure, e.g. >>compiler flags. > > > Configure options are set in the build-configuration files (the defaults are > perlcurrent.cfg [5.9 and 5.8] and perl562.cfg). Merijn has documented them > pretty well and there is some more on that in the FAQ file. Ok, thanks for the pointer >>I have four Solaris boxes running 2.6, 2.7, 2.8 & 2.9, all patched up and >>ready to go. I also have a local 'p4 sync'd copy of the 5.6.x, 5.8.x and >>5.10.x branches from the perl repository and I'm in the process of writing >>a script to drive all four machines. I'm intending that the script will >>sync with the repository a couple of times a day, and automatically kick >>off smokes on machines as they become idle. > > F<configsmoke.pl> creates a little shellscript that starts the smokes, you > could put an endless-loop around the > > /usr/bin/perl smokeperl.pl -c "$CFGNAME" $continue $* > smokecurrent.log 2>&1 > > line. (But I'm no shell programmer) It's a little more complicated - I have to sync the repository up first, and only run if something has changed. >> Hopefully once it is running I >>won't have to touch it. > > You might need to keep your eye on it for a bit, I've had trouble with > Solaris-8 running out of processes (although I'm not sure if that is smoke > related). And if we see test failures, you might need to investigate. Indeed. > I'll see if I can manage some sort of archive mechanism that archives the > report and the smokeperl logfile so one can go back to them whilst in > continous smoke-mode. That would be useful. Another RFE is for a 'status' command that shows you how many tests of a run have completed. -- Alan Burlison --