Re: [pcp] python QA/707 failures

[Ken McDonell <kenj@xxxxxxxxxxxxxxxx>]

On 25/03/13 10:29, Nathan Scott wrote:
> ----- Original Message -----
> > Another multi-platform failure case ...
> >
> > FAIL: test_context (__main__.TestSequenceFunctions)
> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >     File "/home/kenj/src/pcp/qa/src/test_pcp.python", line 355, in
> > test_context
> >       test_pcp(self)
> >     File "/home/kenj/src/pcp/qa/src/test_pcp.python", line 163, in
> >     test_pcp
> >       self.assertTrue(len(inst) >= 2 and len(name) >= 2)
> > AssertionError: False is not true
> >
> > Again clues would be appreciated.
> This again looks like it might be related to running on a single CPU
> machine (test appears wrong).  Its asserting there will be 2 or more
> instances after doing a pmGetInDom on kernel.percpu.cpu.user - if it
> is a single CPU machine then there wont be.
>
> Changing the 2 to "ncpu" (which we happen to have evaluated earlier)
> and the >= to == for both cases, this (still) passes for me - might
> do the trick.
>
> Looks like test 707 also needs an archive-mode invocation Stan?  The
> python test code seems to be wanting one ... but we never invoke it
> with one (AFAICS) and so never go down the archive code paths?  There
> are a bunch of archives in qa/src - would one of those suit?
>
> The test also does " + '.index'" - that should not be required.  Oh,
> its just trying to generate an exception if archive not found?  I'd
> have expected pmNewContext would generate that exception though?
I've addressed the live mode failure, by using sample.bin instead of 
kernel.cpu.percpu.user and making some other changes to fix the Python 
logic that appeared wrong (like calling pmFreeResult and then using 
result) [warning this the first Python code I've written/changed ... be 
very afraid].
Commit coming soon.

I'll leave the archive mode testing to someone else.