[mgp-users 00866] Re: Problems with latest CVS

[David Ronis <ronis@ronispc.chem.mcgill.ca>]

Just an update on the current problems:

I'd written:
>  
>  1.  Reading png images.  I get the message "Too many IDAT's", which is
>      seems to be a warning generated by the png library, followed by a
>      cannot read image msg and exit.  I get it in other programs I use
>      to display the image (e.g., xv), but the image displays.  I played
>      around with a debugger a bit and found that the problem is the
>      call:
>  
>  	png_read_end(png_ptr, info_ptr);
>  
>      at line 217 of image/png.c.  It seems that this version of libpng
>      includes the call to png_read_end() in png_read_png(); in any
>      event commenting out the call fixes the problem.

This fix seems to work.

>  2.  I have several jpeg images to show (essentially one per page) and
>      my mgp file contains lines like:
>  
>  %center, image "Li.jpeg" 256 40 35 1
>   
>      The current version dies with the error:  "Pipeline controller messed up"
>      I've got version 6.2 of libjpeg installed.

I commented out the lines:

    if (input_cols != output_cols || input_rows != output_rows) /* DEBUG */
    ERREXIT(cinfo->emethods, "Pipeline controller messed up");

in image/jpeg.c:fullsize_unsubsample(). This of course, removes the
crash, but doesn't seem to cause other problems; i.e., the
presentation looks right.  


>  
>  3.  I also try and display equations, esentially, using %filter to pass groff 
>  code to groff and to gs as:
>  
>  groff -e $tmp.eqn | gs -sDEVICE=epswrite -sOutputFile=$tmp.eps -q -dNOPAUSE -

This is a gs problem--a bug with epswrite in gs-7.04.  Downgrading to
6.5 fixes it.

David

P.S., Can someone confirm if/when/how this can be commited to CVS?