[gdal-dev] PDF Creation floating point exception for DEFLATE option?

[Even Rouault]

Le lundi 26 novembre 2012 20:59:23, Robert Zermeno a écrit :
> GDAL Community,
>  
> My application generates a geospatial PDF file from a pre-loaded image
> source.  When generating an image, I keep receiving an exception of:
> "exception class EInvalidOp with message 'Invalid floating point
> operation'".  I am using the respoitory from 11-20-2012.  I am using
> Borland C compiler (now known as Embarcadero Codegear). 
> For Borland, I found a solution at
> http://digital.ni.com/public.nsf/allkb/10096EA5DC20FC158625773800628E8A th
> at quited the Invalid floating point operation, but I still get an
> exception. 
> Finally, I built a simple command line program in MSVC 2005 to step through
> GDAL source code to find the exception. 
> The exception occurs when using DEFLATE compression option.  The output
> message I get is located in trees.c (frmts\zlib) from gen_bitlen(s,desc)
> function lines 532-559.  Zlib is used in pdfcreatecopy.cpp Lines 3496-3501
> and 3548.  Line 3548 is where it calls VSIFWriteL(). 
> The output string received is:
>  
> "bit length overflow   
> code 0 bits 6->7"
>  
> or
>  
> "bit length overflow   
> code 6 bits 6->7
> code 0 bits 6->7"
>  
> The code and bits are not always the same.  For me, it did not matter which
> file I used, it always gave me an overflow notice. Can anyone or has
> anyone experienced this issue?  The file still generates but my
> application is halted and a message is presented of the exception. 

You can ignore the "big length overflow" warnings. They are emitted on stderr. 
You only observe them if GDAL is built with -DDEBUG defined, but they should be 
harmless

If you have exceptions/crashes, it is something else.

> Robert