New submission from Guido van Rossum gu...@python.org:
Following my own (Twitter) advice I downloaded the 2.6.4rc1 release
candidate and installed it on my x86 iMac running OS X 10.5.8, and tried
to run a Google App Engine app with it. Google App Engine prefers Python
2.5, but so far it has
Guido van Rossum gu...@python.org added the comment:
FWIW, I've ruled out App Engine SDK 1.2.6 as the source of the regression;
on a MacBook Pro with the same OS 10.5.8 I get a similar traceback with
SDK 1.2.6.
--
___
Python tracker
Guido van Rossum gu...@python.org added the comment:
The full traceback had probably 1000 lines (or whatever the limit is); the
piece between two calls to logging.warn() is repeated over and over.
--
___
Python tracker rep...@bugs.python.org
Philip Jenvey pjen...@underboss.org added the comment:
App engine shows up after every import statement, so it must have some
kind of import hook -- which can do evil things
--
nosy: +pjenvey
___
Python tracker rep...@bugs.python.org
Changes by Philip Jenvey pjen...@underboss.org:
--
nosy: +vinay.sajip
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue7120
___
___
Python-bugs-list
Guido van Rossum gu...@python.org added the comment:
Yes, App Engine uses a PEP 302 style import hook to implement the sandbox.
Could it be that there's a new module dependency introduced by 2.6.3-4?
--
___
Python tracker rep...@bugs.python.org
Benjamin Peterson benja...@python.org added the comment:
While I am not able to reproduce this by running 1.2.6 on Gentoo Linux
with Rietveld, I will make an educated guess that it is the from
multiprocessing import current_process line in r75130. I doubt this
module is available on app engine,
Guido van Rossum gu...@python.org added the comment:
The dependency of logging on multiprocessing feels backwards. But it's
not actually a new regression, it seems that was there in 2.6.2.
--
___
Python tracker rep...@bugs.python.org