Hi Thorsten, *, On Thu, Mar 15, 2007 at 01:45:07PM +0100, Thorsten Ziehm wrote: > > Jogi isn't kidding. Currently there isn't a rule for build-breaker on > tinderboxes.
This was not only a breaker on a tinderbox, it was a breaker that was reported early. Whether it was on tinderbox or not doesn't matter in that case. > This has to be defined, which isn't done until now. We > got a new tooling in EIS The tooling itself is not new, what is (relatively) new is that the results are displayed within EIS for convenience. > and the QA does not know, what to do, when > a one platform is red. > > Do you know, if the new tooling in EIS works without any error? Only > then it is possible to define a rule. You didn't read carefully enough. (And that shows again why ToFu is bad and basiaclly a waste of time): ,----[ I wrote ] | If you want to say: QA itself should not be bothered to check tinderbox | status themselves, then fine. But: I already checked the status for you | and told you it breaks. `---- So: QA didn't need to check themselves I already checked and reported the breakage. And: Yes, I know that the tooling works. I maintain three buildslaves (one of them is being relocated right now and thus not available right now). The false positives "reds" occur rarely, but can easily be spotted. False positive "greens" don't happen. > I saw many CWS in the last weeks and only some of them are green on > all platforms. But I couldn't found one, without an error. You surely didn't look at the overview page that lists all the ready-for-QA cws: http://go-oo.org/tinderbox/ready_for_QA.express.html Here: Basically everything is green except from: freeafms (new module, is not passed by EIS and thus breaks, can be ignored) npower6 changed configure and helpers, maybe not generated properly. hr40 - java crashed, but is already built again by the buildslave the "linux-gcc-4.0.3 (gij)" box is not in use anymore, so its status can be ignored (only long-lived cws can show it, since it is not in use for quite a while now) > So what should be the rules to reject a CWS, when 50 errors are shown, > but the color is green, or when 20 errors are shown and it is violet > or when 10 errors exists and it is red. Simple (and that has been mentioned in the announcement): If it is green, you don't need to have a look at all. If it is read: Visit the tinderbox page and have a look at the summary logs. Any false positive breakers can be easily identified: * Problems with anoncvs * new modules * something in configure was messed with, but not updated Every thing else should deserve a closer look. One other case where the break can be ignored is when the break appears in the master as well (and can only happen on the newest milestones[1], so having a look at the global "ready-for-QA" page that also lists the last three masters will tell you whether that is the case) http://go-oo.org/tinderbox/ready_for_QA.express.html > Only clear rules will help to eliminate these problems. But if a rule > is defined and exist, all CWS has to be handled like this. > > Currently I do not know, what the numbers and colors should tell me. > A clear definition will help here. Simple: http://www.mail-archive.com/[email protected]/msg03936.html > [fullquote snipped] ciao Christian [1] since admins of the buildslaves will take care of collecting and applying the necessary patches then -- NP: Slipknot - Iowa Join #qa.OpenOffice.org on irc.freenode.net --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
