[pygtk] Strange crash issue on Windows w/ PyGTK, Cairo...

[CJ Kucera]

(redirected this back on-list, I assume it was an inadvertant
non-group-reply)

David Ripton wrote:
> On 2009.03.18 15:22:12 -0500, CJ Kucera wrote:
> > Certainly does - it's nice to see that someone else has at least seen
> > that kind of crash before.  I'm still incredibly mystified by it.  I
> > just don't see how adding a "pass" statement to a random function can
> > mean the difference between a crash and a working app.  Ditto for
> > removing unused functions, etc...
> 
> I don't either.  I think you're doing something else wrong.
> 
> Run your top-level script with python -v, and triple-check all the
> import paths.

Thanks for that suggestion - I did just try with "python -v" and
everything looks kosher there.  Nothing from outside of C:\Python26\lib
or C:\Python26\lib\site-packages is getting loaded (apart from my own
classes, of course), and it's all stuff I've just recently installed.
These Windows images don't get much use, so there's really no cruft on
them.

In terms of doing something wrong, certainly something in my code is
triggering this, though IMO I'd consider this behavior a bug of some
part of Python (whether it's the interpreter itself or the pygtk/pycairo
bindings, etc) or gtk+ itself.  One of the benefits of using an
interpreted language like Python is generally that you don't have to
deal with this segfault-like behavior.  If something goes wrong, I
should get a Python exception that I can catch and deal with.  Also I
feel that my code isn't really "out there" - after all, it does work
perfectly fine in Linux, and during the first compilation stage on
Windows, even.

I have made some good progress in paring the program down, though.  I'm
hoping to be able to get a nice, small reproducible script pretty soon.
One of the blocks that I tried deleting actually flipped around the
error pattern to be exactly the opposite of before.  *Now* I get that
Windows exception on the very FIRST time through, and every subsequent
time seems to run fine.  At any rate, now at least the problem seems to
be less subject to apparent whimsy, and I may have isolated it to a
Cairo function which is reading in PNG data from a string.  Hopefully
I'll have it fully isolated in a bit, and I'll be able to post some
actual code.

Thanks for the reply!

-CJ