Make evaluateCommand reuse parent's result where needed....

Make evaluateCommand reuse parent's result where needed.

Also use dir() to check if evaluateCommand is defined
on the log processor object.

BUG=18597
TEST=PerformanceGTestLogProcessor returns correct
result code.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=25546

[chase, 2009-09-04 19:09:50]

I verified this still works on my local system for the main waterfall/linux perf
+ perf_expectations.

[Nicolas Sylvain, 2009-09-04 21:05:42]

http://codereview.chromium.org/194029/diff/1/2
File chromium_step.py (right):

http://codereview.chromium.org/194029/diff/1/2#newcode130
Line 130: return self._log_processor.evaluateCommand(cmd)
Maybe here we should do  :
<pseudo code>
if self._lo...evaluateCommand is SUCCESS:
  return ShellCommand....EvaluateCommand.
</pseudo code>

We don't want to force it to SUCCESS if the return code is non-zero.

http://codereview.chromium.org/194029/diff/1/2#newcode133
Line 133: return shell.ShellCommand.evaluateCommand(cmd)
hmm, are you sure that works? I'm afraid it will create a new instance of
ShellCommand instead of using the one we inherit from.  Can you do something
like, return super(ProcessLogShellStep, self).evaluateCommand(cmd) ?

[Nicolas Sylvain, 2009-09-04 21:06:53]

http://codereview.chromium.org/194029/diff/1/2
File chromium_step.py (right):

http://codereview.chromium.org/194029/diff/1/2#newcode133
Line 133: return shell.ShellCommand.evaluateCommand(cmd)
Actually. ShellCommand.evaluateCommand(self, cmd) might work too. just need to
make sure it has self.

On 2009/09/04 21:05:42, Nicolas Sylvain wrote:
> hmm, are you sure that works? I'm afraid it will create a new instance of
> ShellCommand instead of using the one we inherit from.  Can you do something
> like, return super(ProcessLogShellStep, self).evaluateCommand(cmd) ?
>

[chase, 2009-09-04 22:28:18]

Linux reference build is crashing locally so I'm unable to verify this patch
end-to-end on the log processor code path.  Should work, though.  Let me know if
you have more feedback.

http://codereview.chromium.org/194029/diff/1/2
File chromium_step.py (right):

http://codereview.chromium.org/194029/diff/1/2#newcode130
Line 130: return self._log_processor.evaluateCommand(cmd)
On 2009/09/04 21:05:42, Nicolas Sylvain wrote:
> Maybe here we should do  :
> <pseudo code>
> if self._lo...evaluateCommand is SUCCESS:
>   return ShellCommand....EvaluateCommand.
> </pseudo code>
> 
> We don't want to force it to SUCCESS if the return code is non-zero.

Done.

http://codereview.chromium.org/194029/diff/1/2#newcode133
Line 133: return shell.ShellCommand.evaluateCommand(cmd)
Got it, done.

[Nicolas Sylvain, 2009-09-04 23:02:08]

LGTM