[Buildbot-devel] ShellCommand timeout when command fails (win32)

[Simo Salminen]

On Thu, Feb 24, 2005 at 10:36:41PM -0800, Brian Warner wrote:
> > I have the following build step in BuildFactory:
> > s(step.Test, command=["check.bat"])
> > 
> > However, in my situation, the file check.bat is not found, and the slave
> > correctly complains about that (see below). The problem is that the
> > buildmaster doesn't take notice of this, but waits for the timeout.
> > I would expect it would report failure immediately. Is there any
> > reason why it should wait in the timeout when the command clearly
> > fails?
> 
> Hrm. You're right. It looks like the slave noticed the problem (the exception
> was caught by bot.SlaveBuilder.commandComplete) and sent back an error
> message, but something is broken with that error-reporting mechanism, as it
> didn't cause the step to be flunked.
> 

Uh, now it works fine, cannot reproduce.

Now the slave log says:
<- clip ->
2005/02/26 21:02 FLE Standard Time [Broker,client]
SlaveBuilder.commandFailed <b
uildbot.slave.commands.SlaveShellCommand instance at 0x01025328>
2005/02/26 21:02 FLE Standard Time [Broker,client] Failure:
buildbot.slave.comma
nds.AbandonChain: -1
<- clip ->


This 'AbandonChain' didn't happen before, IIRC. Weird.


When I had the problem, the master side log said this:
<- clip ->
2005/02/22 23:59 EET [-] ShellCommand.start using log
<buildbot.status.builder.LogFile instance at 0x4082ea8c>
2005/02/22 23:59 EET [-]  for cmd <RemoteShellCommand '['check.bat']'>
2005/02/22 23:59 EET [-] <RemoteShellCommand '['check.bat']'>:
RemoteCommand.run [16]
2005/02/22 23:59 EET [-] command '['check.bat']' in dir 'build'
2005/02/22 23:59 EET [-] LoggedRemoteCommand.start
<buildbot.status.builder.LogFile instance at 0x4082ea8c>
2005/02/23 00:03 EET [Broker,8,193.166.235.60] <Builder
'foo'>.detached
<- clip ->

So nothing reported until I closed down the slave. I'll report if I
manage to get this happen again.

-- 
Simo Salminen