[Buildbot-devel] buildbot stop, buildbot start results in corrupted waterfall disp lay

[Elliot Murphy]

Hello,

I think I've run into a persistence problem with the buildbot master.

 

Using the latest source from CVS and running on Linux with python 2.3.4,
I've created a very simple buildmaster with two slaves (A and B). After
starting the buildmaster, I can view the HTML waterfall display, and the
display looks correct, with a column for each buildslave.

 

I then run buildbot stop followed by buildbot start on the master.
Refreshing the HTML waterfall display, I see the status updated to show that
the master shutdown and started again. However, both builder columns now
show slave B, and both links go to the ping page for slave B. If I manually
enter the URL for the ping page for slave A, it says "Builder: B", NOT
CONNECTED (slave 'A').

 

If I stop the master, remove changes.pck, and restart the master, the
display is still corrupted. If I stop the master, remove both changes.pck
and the "full" directory, all the history is lost but the display shows both
build slaves correctly, not two instances of slave B.

 

It seems like something is being corrupted when the buildbot master is shut
down, but I'm not sure how to track it down further.

 

Thanks for your help,

-elliot

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://buildbot.net/pipermail/devel/attachments/20041111/943c7306/attachment.html>