selftest results for lp:bzr revno.4804

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vincent Ladeuil wrote:
>>>>>> "jam" == John Arbash Meinel <john at arbash-meinel.com> writes:
> 
> <snip/>
> 
>     jam> We have a bit of kipple from:
>     jam> Exception RuntimeError: 'maximum recursion depth exceeded in
>     jam> __subclasscheck__' in <type 'exceptions.RuntimeError'> ignored
> 
> That one is present almost everywhere.

Sure, most of the remaining 'kipple' is present everywhere and unrelated
to win32. Instead, it is mostly related to... we don't run the test
suite manually enough to see it, to annoy us into fixing it.


> 
> <snip/>
> 
>     jam> Seemed to fix them for me, but that is a kludgey
>     jam> workaround. I think the issue is that we aren't
>     jam> synchronously closing the file stream over sftp (instead
>     jam> doing it asynchronously). And on a single CPU machine,
>     jam> the sftp thread never gets control to actually run
>     jam> 'close'. Hence why 'time.sleep(0.001)' is enough to make
>     jam> it work.
> 
> That sounds exactly like the sync bugs I'm fixing to avoid leaks,
> sleeping is another way to transfer control to the other thread
> when the threads aren't synchronized. It may still fail but more
> rarely.
> 
>     jam> I should note for Vincent that I never run with
>     jam> --parallel.
> 
> Hmm, well, that means the sync bugs manifest themselves
> differently on windows and the leaks have different consequences.
> 
> By the way, how many leaking tests are reported ?
> 

I think --subunit supresses the leaking test message. At least my last
run didn't seem to say anything about them. And currently I use
- --subunit whenever I run the *full* test suite, because it lets me
easily | tee  the result, and I prefer doing that when the test suite
takes >1hr to complete.


>     jam> I *do* tend to run with --subunit, but it isn't required
>     jam> for the test suite to pass.
> 
> Good, a different result would have been annoying.
> 
>       Vincent
> 

Well, there are some differences for debugging purposes, etc. But I
don't think I've seen test correctness differences. :)

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAksX47MACgkQJdeBCYSNAAM8EACghl/zleqdYpr8R3KvOivlGqhp
3UgAn2B3cZ4zObYPwXKWpXIKNVScs5pH
=nalw
-----END PGP SIGNATURE-----