[Buildbot-devel] Extending ShellCommand

[Kirill Lapshin]

Hello,

I just started playing with BuildBot, so want to apologize in advance 
for potential FAQ.

First thing I want to congratulate developers for such a nice product. 
It does require some polish, but even it is current state it is very 
usable and has quite a lot of nice features missing in CruiseControl. 
Thanks a lot!

I am trying to extend ShellCommand to run unittests and update status 
nicely. Running unit tests works just fine, but updating status throws 
an exception I can't decipher.

Here is relevant code:

class RunUnitTests(ShellCommand):
    name = 'unittests'
    def __init__(self, **kwargs):
        ShellCommand.__init__(self, **kwargs)
        self.python = kwargs['python']

    def start(self):
        command = self.python + ' '
        command += os.path.join('cmd', 'test_all.py')

        self.command = command
        ShellCommand.start(self)

    def startStatus(self):
        verb = "testing"
        self.setCurrentActivity(Event("yellow", [verb],
                                      files={'log': self.log}))
    def finishStatus(self, result):
        verb = "unittests"
        if type(result) == types.TupleType:
            result, text_ignored = result
        if result == FAILURE:
            self.updateCurrentActivity(color="red", text=[verb, "failed"])
        else:
            self.updateCurrentActivity(color="green", text=[verb])
        self.finishStatusSummary()
        self.finishCurrentActivity()


What happens is startStatus() updates status to yellow 'testing' as 
expected, log link works just fine, but when step is done, whole build 
gets marked as successful, while unittests step still remains in yellow 
'testing' mode. Hitting on log link at this point genereates following 
traceback:


2004/07/06 21:16 EDT [HTTPChannel,0,127.0.0.1] Traceback (most recent 
call last):
          File 
"/usr/lib/python2.3/site-packages/twisted/protocols/basic.py", line 223, 
in dataReceived
            why = self.lineReceived(line)
          File 
"/usr/lib/python2.3/site-packages/twisted/protocols/http.py", line 965, 
in lineReceived
            self.allContentReceived()
          File 
"/usr/lib/python2.3/site-packages/twisted/protocols/http.py", line 1006, 
in allContentReceived
            req.requestReceived(command, path, version)
          File 
"/usr/lib/python2.3/site-packages/twisted/protocols/http.py", line 557, 
in requestReceived
            self.process()
        --- <exception caught here> ---
          File "/usr/lib/python2.3/site-packages/twisted/web/server.py", 
line 165, in process
            self.render(resrc)
          File "/usr/lib/python2.3/site-packages/twisted/web/server.py", 
line 172, in render
            body = resrc.render(self)
          File 
"/usr/lib/python2.3/site-packages/buildbot/status/event.py", line 384, 
in render
            data += self.content(self.getEntries(), asText)
          File 
"/usr/lib/python2.3/site-packages/buildbot/status/event.py", line 356, 
in content
            for type, entry in entries:
        exceptions.TypeError: iteration over non-sequence


on line 356 of event.py entries is already None. Any ideas?

I am using buildbot 0.4.3 on Debian.

On a related note, buildbot did not like a couple of our unittests. The 
whole unittesting process was dying saying that it caught signal 1. The 
very same tests run just fin on the same machine, run by the same user 
from the same folder. On closer examination it turned out that gnuplot 
related tests were to blame. They did not do anything special, just 
tested usage of Gnuplot.py, which basically spawns gnuplot process and 
pipes some data into it. Any ideas why tests may run fine from shell, 
but not from buildbot?

--Kirill