[Image-SIG] Debugging PIL in Win32?

[Vladimir Pastukhov]

Hello,

Brendler, Ralph wrote:
> Thanks for the input.  I anxiously await the new library!  Any idea when this will become part of the standard PIL release?

This depends on PIL guys.  Currently I am using PIL 1.1.5-alpha
for development and going to make it ready for the 1.1.5 release.

> I deal mostly with TIFFs from the US Patent Office (all group4 encoded), so I was thrilled when you added this support to PIL.  I was hoping to repay you by tracking down this segfault, but this one has me completely stymied.
> 
> I have validated the compilation flags on all of my libraries, and everything seems to be correct, but I am still getting the heap violation from libTiff.  This one really has me stumped.  It looks like *someone* is setting DEBUG_NEW (adds some marker bytes around the allocation to flag memory bugs), but I can't find this anywhere.

What C compiler are you using?  I've built it with Visual C++ 6
without any special settings.

> Any other suggestions?  I am moving forward with my development on Linux, but eventually I will need this to work on Win32...

I've attached Win32 binary to a separate letter, so far had no
problems with it.

> 
> I appreciate your help.
> 
> -----Original Message-----
> From: Vladimir Pastukhov [mailto:vpastukhov at naumen.ru]
> Sent: Thursday, April 22, 2004 5:10 PM
> To: Brendler, Ralph
> Cc: image-sig at python.org
> Subject: Re: [Image-SIG] Debugging PIL in Win32?
> 
> 
> Hello,
> 
> Brendler, Ralph wrote:
> 
>>The PIL code is straight 1.1.4 plus the July patch for Group4 support.
> 
> 
> BTW, next month I plan to release a new patch that will not use
> internal libtiff structures access (requires some additions to libtiff
> API, which are settling). This should resolve build and compatibility
> problems as yours.
> 
> 
>>As far as I can tell everything appears to be compiled with the same options, but I am getting an heap validation error (which usually indicates a debug/nondebug difference in libraries) when calling ImagingLibTiffCleanup.
> 
> 
> Looks like different optimizations are used for the libImaging library
> and for the Python extension itself.  Try checking CFLAGS in libImaging/
> Makefile.win and the options setup.py uses to compile decode.c
> 


-- 
Vladimir Pastukhov <vpastukhov at naumen.ru>