2009.04.07 14:13 "Re: [Tiff] Writing tiff to in-memory buffer", by Edward Lam

Fixing that gives me a valid TIFF*. I also need to setup my seek_proc to prohibit it from seeking on an empty ostream. Doing though would raise the failbit. I don't understand why TIFFClientOpen is calling seek_proc( 0,0). It seems redundant to me.

This was noted a while back when tif_stream.cxx was tested (see http://www.asmail.be/msg0055045547.html). It was thought to be a bug at the time, but then later resolved to be correct. Unfortunately, I can no longer find the subsequent discussion that resolved it in libtiff bug database. (I forget whether the bug was logged before the libtiff bugdb migration.)

Here is the current comment from the trunk source:

/*

• The doc for "fopen" for some STD_C_LIBs says that if you
• open a file for modify ("+"), then you must fseek (or
• fflush?) between any freads and fwrites. This is not
• necessary on most systems, but has been shown to be needed
• on Solaris.

  */
TIFFSeekFile( tif, 0, SEEK_SET );

Regards,

-Edward