On Oct 30, 2007, at 1:16 PM, Heikki Toivonen wrote:
bear wrote:
D John Anderson wrote:
Unfortunately, the new recorded scripts tests are catching bugs in
rt.py and the tinderboxes themselves (random false positives and
false
negatives) that look unrelated to recorded scripts, so I'm blocked
until I can get some help fixing them.
That is yet unproven - rt.py and tinderbox are running the tests the
exact same way in both trunk and the recorded_scripts branch and
we are
giving you help in finding out what is wrong in so far as we can
with rt
and tinderbox.
I'll say it again - rt.py just reports what the return code from
Chandler is and dumps the log files. Tinderbox does exactly the same
from rt.py.
As far as I can tell, rt and Tinderbox are working correctly. However,
bear pointed out that Tinderbox is doing |make distrib| while most
developers are doing |make install|. It was quite a surprise to see
that
when I did |make install| the recorded scripts failed, but succeeded
with |make distrib|.
The only new bugs in the process were to the Chandler launcher where
some environment variables were shown to need changes and one of
those
has been made and the other will be pushed to trunk today.
So assuming my find is correct, would it be enough to just update
the rs
branch? That should get the rs branch using the Chandler.po like it
should.
Or is there something more we want to do about make distrib target?
It would be nice to address the problem with the distrib target
before committing the Chandler.po fix -- otherwise there would be no
way to verify that the make distrib problem is really fixed, since
the Chandler.po fix would mask it.
--
Heikki Toivonen
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
