AWARE SYSTEMS
TIFF and LibTiff Mail List Archive

Thread

2001.12.12 20:23 "jpeg in tiff problem", by Eric Hamrick
2001.12.13 02:03 "Re: jpeg in tiff problem", by Frank Warmerdam
2001.12.14 01:42 "Re: jpeg in tiff problem", by Tom Lane
2001.12.14 14:00 "Re: jpeg in tiff problem", by Eric Hamrick
2001.12.14 14:23 "Re: jpeg in tiff problem", by Frank Warmerdam

2001.12.14 01:42 "Re: jpeg in tiff problem", by Tom Lane

The file doesn't have a YCbCr sampling value in the TIFF directory, so the library defaults to 2,2. However, the header of the jpeg data stream indicates a 2,1 sampling. Thus the error.

Sigh... another implementor who didn't read the spec, and couldn't be bothered to do any cross-testing.

I tried overriding the default sampling value with the value in the jpeg data stream and that seemed to work, so I have modified tif_jpeg.c to always do this if they differ, and to just issue a warning instead of an error.

Seems a reasonable hack. But no doubt someday we'll see another broken implementation where the opposite "fix" is needed to read its output. I trust the warning is not easily silenced?

In the meantime, nastygrams to imagegear are definitely in order. If they want to argue that their code conforms to TTN#2, tell 'em to take it up with me.

regards, tom lane
organizer, Independent JPEG Group