Found the problem with Python 2.6 and subprocess. Python 2.6 is looking for the environment variable "SystemRoot", but doesn't find it. I ran the Pygame unit tests with the necessary fix and it worked. Everything but the numpy dependent tests passed. I will not commit the change to SVN because I still consider it a Python 2.6 bug. It breaks existing code and is Windows specific.
Lenard
On Jul 23, 2008, renesd@xxxxxxxxx wrote:
awesome work!!!
On Thu, Jul 24, 2008 at 4:24 AM, <len-l@xxxxxxxxx> wrote:
> Sorry, make that bug fixed in Revision 1530.
>
>
>
> As for the run_test.py -s option, that appears to be a Python 2.6 bug
> concerning environment variable passing. The following program crashes with
> the error popup dialog "The application failed to initialize properly
> (0xc0150004)":
>
>
>
>
>
> import subprocess
>
> command = '''python -c "import os; print 'YAHOO=%s' %
> os.environ['YAHOO']"'''
>
> subprocess.Popen(command, env={"YAHOO": "ABCDE"}).wait()
>
>
>
>
>
> I will submit a bug report to the Python issue tracker. Otherwise things
> look promising with Python 2.6. surfarray and sndarray remain untest as
> there is not numpy support for 2.6 yet.
>
>
>
> Lenard
>
>
>
>
>
>
>
> On Jul 23, 2008, len-l@xxxxxxxxx wrote:
>
> Revision 1520 fixes bug that crashed Python 2.6 in transform_test.py.
>
>
>
> Lenard
>
>
>
>
>
> On Jul 22, 2008, len-l@xxxxxxxxx wrote:
>
> Pygame is linked to msvcr90.dll for Python 2.6. So data mixing shouldn't be
> the problem. Also, other transform tests pass. Just
> test_threshold__uneven_colors page faults. I have been playing with
> surf_threshold in transform.c. I find removing all calls to SDL_MapRGBA,
> even those not actually executed by the unit test, stops the page faults. I
> am at a loss right now. The only other Python 2.6 problem I found is with
> the run_tests.py subprocess option. Python 2.6 doesn't like running itself
> as a subprocess. It is probably not worth delaying the 1.8.1 release over.
> If I figure this out then I can make the Python 2.6 Pygame binary available
> then.
>
>
>
> Lenard
>
>
>
>
>
> On Jul 22, 2008, renesd@xxxxxxxxx wrote:
>
> Maybe it is because it's using malloc and free in there?
>
> It shouldn't mess with things... because that data is never mixed with
> python memory... but maybe that's it.
>
> That's my only guess at the moment.
>
>
> On Wed, Jul 23, 2008 at 5:34 AM, <len-l@xxxxxxxxx> wrote:
>> Hi,
>>
>
> [snip]
>
>> If I can figure out what is going on with pygame.transform I may also
>> get a Python 2.6 release binary out.
>>
>>
>>
>> L. L.
>>
>>
>>
>>
>>
>> On Jul 22, 2008, renesd@xxxxxxxxx wrote:
>>
>> hi,
>>
>> pygame 1.8.1rc3 was marked in subversion.
>>
>> What's new can be found here: http://pygame.org/whatsnew.shtml
>>
>> http://rene.f0o.com/~rene/stuff/pygame-1.8.1rc3.zip 1.7MB
>> d148d890ef7ba8a07804718a0c9cad1d
>>
>> http://rene.f0o.com/~rene/stuff/pygame-1.8.1rc3.tar.gz 1.5MB
>> 03d078502aeec6ee1aea31c7d0c74c4c
>>
>> You can get automated builds for win/mac from here:
>> http://thorbrian.com/pygame/builds.php
>>
>> There's only been one change since Lenards last windows binary release.
>>
>> I think this is definitely it for 1.8.1.
>>
>> Unless there are patches for bugs with unittests, or a really bad bug
>> is found, this will be 1.8.1 to be released this weekend.
>>
>> This will likely become 1.8.1 with hopefully no changes.
>>
>>
>> cu,
>>
>