[Buildbot-devel] console does not match waterfall for older builds (0.7.12 + CSS fix; SVN)

Nicolas Sylvain nsylvain at chromium.org
Mon Feb 8 01:48:09 UTC 2010

For some reasons the console does not look before build 85 for the FreeBSD
build. So it thinks
it saw r2317 first in build 85, so it attributes the failures in b85 to that
revisions as well.  (while it fact it should have seen it first in build

The problem also seem to happen for the other builders.

By looking at the debug log at the bottom, it looks like it is using only 10
builds results to build the whole console, which is
clearly not enough. (and not expected)

If you are familiar with python and want to debug, I would look into
in buildbot/status/web/console.py

In theory this function is supposed to return an array of all the builds
needed to compute the console for a given builder. You
can watch for calls to this function with "builderName" = "FreeBSD-i686",
and then you can add output to see what's going on
in the while.

This function is supposed to look at 40 builds maximum, but it seems to stop
early. There are 2 reasons:

1.  Maybe the self.comparator.isRevisionEarlier function is failing. Maybe
it's returning true when it should
not. I would try adding some prints there.

2. getPreviousBuild() return None.  This is outside the scope of the
console, but I doubt this is it.

My bet so far is that isRevisionEarlier is not called correctly. I thought
we fixed that though.

Can you replace :

                # Now break if we have enough builds.
                current_revision = self.getChangeForBuild(
                    builder.getBuild(-1), revision)
                if self.comparator.isRevisionEarlier(
                    devBuild, current_revision):

               # Now break if we have enough builds.
               if int(got_rev) < int(revision):

And try again. Maybe it will work. (you need to restart the master).

If you can't, and your build stuff is not too secret, you can also zip your
master directory and send
it to me so I can try to debug on my machine.

Hope this helps


On Sun, Feb 7, 2010 at 3:15 PM, Charles Lepple <clepple at gmail.com> wrote:

> Before I file a ticket, I just wanted to run this by the list to make
> sure I am not misinterpreting one of the existing issues.
> I am testing the console view on 0.7.12 (with the patched CSS file),
> and I am running Buildbot against a Subversion repository.
> The waterfall view reflects the actual state of the builds:
>   http://www.ghz.cc/charles/tmp/2010-02-07/nut-waterfall.png
> But the console view doesn't always seem to keep the state for old builds:
>   http://www.ghz.cc/charles/tmp/2010-02-07/nut-console.png
> Specifically, r2317 passed for all builders, and 2318-19 failed for
> all but the left builder.
> I don't see anything odd in twistd.log.
> Any ideas on where to look to debug this?
> --
> - Charles Lepple
> ------------------------------------------------------------------------------
> The Planet: dedicated and managed hosting, cloud storage, colocation
> Stay online with enterprise data centers and the best network in the
> business
> Choose flexible plans and management services without long-term contracts
> Personal 24x7 support from experience hosting pros just a phone call away.
> http://p.sf.net/sfu/theplanet-com
> _______________________________________________
> Buildbot-devel mailing list
> Buildbot-devel at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/buildbot-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://buildbot.net/pipermail/devel/attachments/20100207/79f85fda/attachment.html>

More information about the devel mailing list