[Python-Dev] RE: test_sax failing (Windows)

[Tim Peters]

[Martin, on ftp://ftp.jclark.com/pub/xml/expat.zip]
> ...
> That gives some clue, yes. Unfortunately, that URL itself is a symlink
> that was expat1_1.zip (157936 bytes) at some point,

That's the one I've been using.

> and now is expat1_2.zip (153591 bytes).

I'm assuming you're recommending that one!  Based on that assumption, I've
downloaded a new one and will put that in the 2.1a1 Windows release.  Scream
if that's not what you want.

> ...
> Anyway, I was using 1.1 in my own tests, and 1.2 in PyXML - either
> works for me. I never tested 1.95.x (which is also not available from
> jclark.com).

If you do and love it, let me know where to get it and I'll ship that
instead.

>> The tests passed in the wee hours (EST; UTC -0500) this morning.
>> They began failing after I updated around 1pm EST today.

> I just merged pyexpat changes from PyXML into Python 2 so that could
> be the cause. However, this very code has been used for some time by
> PyXML users, why it crashes for you is a mystery to me.

Perhaps gc, perhaps uninitialized vars, ..., hard to say.  Unfortunately,
it's not unusual for flawed code to display different behavior across
platforms; or, from the long-term QA perspective, it's *great* that flawed
code doesn't always appear to work on all platforms <wink>.

> Any chance of producing a C backtrace?

Sent that before; doesn't look like much help; we're seeing a NULL type
pointer, but at that stage there's no telling when or where or why it
*became* NULL.

I'm going to rebuild the world from scratch, and use the new DLLs.  You
should assume that didn't help unless I say otherwise within 15 minutes.