ImageMagick interprets PGM header incorrectly

[pipe]

Binary package hint: imagemagick

Ubuntu 8.10
imagemagick 7:6.3.7.9.dfsg1-2ubuntu3

When reading a PGM-P5 (binary) image with a maxval of 256, the 'convert' and 'display' utility, and likely the rest, will interpret each pixel as one byte, when it should interpret each pixel as two bytes since 256 is more than 255. This causes vertical stripes and other nastiness in the image. If maxval is set to 257, everything works as it should.

Apparently you can only submit one attachment in this bugtracker so users have to host their own files. Download example images from:
http://natrox.ath.cx/imagemagick/lena-255.pgm
http://natrox.ath.cx/imagemagick/lena-256.pgm
http://natrox.ath.cx/imagemagick/lena-257.pgm

Display with imagemagick:
$ display -sample 400% lena-255.pgm
$ display -sample 400% lena-256.pgm
$ display -sample 400% lena-257.pgm

255 and 257 works, 256 breaks.

[pipe (pipatron)]

A patch that solves this issue:

pipe@vitamin:~/src/imagemagick-6.3.7.9.dfsg1$ diff -u coders/orig/pnm.c coders/pnm.c
--- coders/orig/pnm.c 2007-12-08 01:13:40.000000000 +0100
+++ coders/pnm.c 2009-03-24 21:54:47.000000000 +0100
@@ -417,7 +417,7 @@
       ThrowReaderException(CorruptImageError,"NegativeOrZeroImageSize");
     if (max_value >= 65536)
       ThrowReaderException(CorruptImageError,"ImproperImageHeader");
- for (image->depth=1; (1UL << image->depth) < max_value; image->depth++);
+ for (image->depth=1; (1UL << image->depth) <= max_value; image->depth++);
     scale=(Quantum *) NULL;
     if (image->storage_class == PseudoClass)
       if (AllocateImageColormap(image,image->colors) == MagickFalse)

[Brian Murray (brian-murray)]

Thanks for taking the time to report this bug and including a patch too. I've tried recreating this on the Jackalope using package version 7:6.4.5.4.dfsg1-1ubuntu3 and was unable to. Looking at the source code of imagemagick it seems to have change a bit too:

    if (max_value >= 65536)
      ThrowReaderException(CorruptImageError,"ImproperImageHeader");
    for (depth=1; GetQuantumRange(depth) < max_value; depth++) ;
    image->depth=depth;
    if ((image_info->ping != MagickFalse) && (image_info->number_scenes != 0))
      if (image->scene >= (image_info->scene+image_info->number_scenes-1))

Since you wrote the patch, do you think this might resolve the bug? Thanks in advance.

[Mathew Hodson (mhodson)]

Was this fixed upstream?

[Launchpad Janitor (janitor)]

[Expired for imagemagick (Ubuntu) because there has been no activity for 60 days.]