[Stackless] Python25.dll crashing on Windows 7

[Kristján Valur Jónsson]

I've had a private reply from MS, and it turns out to be an eve-online crash that they are seeing.  My current suspicion is that it is the crash-on-exit bug (swapping in an invalid stack) that I have been chasing for years now.  When I caught that in action, somehow the stack management machinery had gone amok.
Perhaps windows 7 will provide for a more stable envirionment to chase down that problem.

Anyway, the winqual people have found their intended target in CCP Games, and we'll be sorting this out with them from here :)

Cheers,

K

> -----Original Message-----
> From: stackless-bounces at stackless.com [mailto:stackless-
> bounces at stackless.com] On Behalf Of Richard Tew
> Sent: 7. ágúst 2009 08:29
> To: Colleen Williams (Smart Marketing & Events)
> Cc: stackless at stackless.com
> Subject: Re: [Stackless] Python25.dll crashing on Windows 7
> 
> On Fri, Aug 7, 2009 at 9:38 AM, Colleen Williams (Smart Marketing &
> Events)<v-collw at microsoft.com> wrote:
> > Python25.dll is crashing consistently on all Windows 7 builds. You
> may
> > already be aware of the issue but it appears the crash is related to
> a
> > manipulation of stacks. The app manipulates the C stack directly for
> its own
> > implementation of lightweight threads. This means a Stack Buffer
> Overrrun
> > could be coming from almost anything it's doing. This is causing
> almost 4000
> > crashes per day since mid-June. The version we're seeing crashing the
> most
> > is 2.5.4150.1013.
> 
> I've been working on Stackless Python 2.5.3 today, and interestingly,
> I get a stack overflow crash with it.  This crash exists in the
> released Python 2.5.3 code I am working from.  It is pretty much a
> variant of an old Python bug (http://bugs.python.org/issue2286).
> 
> The call stack looks something like this:
> >	python25_d.dll!_PyObject_GC_New(_typeobject * tp=0x104833e0)
> Line 1345 + 0x9 bytes	C
>  	python25_d.dll!PyDict_New()  Line 220 + 0xa bytes	C
>  	python25_d.dll!r_object(WFILE * p=0x0021360c)  Line 835 + 0x5
> bytes	C
>  	python25_d.dll!r_object(WFILE * p=0x0021360c)  Line 842 + 0x9
> bytes	C
>         .........
> 
> Now, this was encountered with VC 2008 express.  The official compiler
> for the 2.5 branches of Python in order to maintain compatibility with
> built extensions is actually Visual Studio.NET 2003, which I only have
> the free toolkit of, and under that the crash is not encountered.
> 
> Again, this is not a Stackless problem, but going by my experience
> with the crash today, if I had to guess where the thousands of
> failures for Python 2.5.4 were coming from, I would assume someone had
> compiled it with VS 2008.  Perhaps a buildbot that was constantly
> rerunning the unit tests, or snakebite (http://snakebite.org).
> 
> Anyway, just some further thoughts.
> 
> Cheers,
> Richard.
> 
> _______________________________________________
> Stackless mailing list
> Stackless at stackless.com
> http://www.stackless.com/mailman/listinfo/stackless