CHANGES to libtiff-3.8.0 : jD Gascuel, 2006-01-11 :

  - tif_config.h for configuration from rules.pri/TiffIO.pri, and
    Added HAVE_INT_TYPES in tiff.h to avoid clash in AIX.
    Removed tiffconf.h (obsolete, *BUT* used too).
  - Fix compile errors (http://bugzilla.remotesensing.org bug 1033).
  - Fix crash (http://bugzilla.remotesensing.org bug 1034).

CHANGES to libtiff-3.7.0 : jD Gascuel, 2004-11-05 :

  - To enable compilation with a <<modern>> compiler (aka MSVC++ 7.1),
    I had to twick a bit the source here and there, eg. adding casts
    on allocated space before store to its pointer.
    
  - Mixed together tif_config.h and tiffconf.h, and twick a bit the
    compilation flags to get support for LZW, TIFF, and JPEG from Qt.

CHANGES to libtiff-3.7.1 : jD Gascuel, 2005-01-14 :

  - Downloaded brand libtiff-3.7.1.zip.

  - Have to redo the same twicks (casting CLAMP() to (uint16), to avoid
    data loss warnings, casting _TIFFmalloc() to the appropriate pointer type,
    changing pow() to powf() in tif_pixarlog.c.

  - Have to fix bugs in tif_win32.c Win32WarningHandler() and 
    Win32ErrorHandler() that messed-up with uninitialized szTmp variable.

  - Have to fix tif_getimage.c TIFFReadRGBAImageOriented() to initialize
    emsg to "no error" in case TIFFRGBAImageBegin() returned an error without
    setting it (eg when reading 16 bits LAB or CYMK).

CHANGES to libtiff-3.7.2 : jD Gascuel, 2005-06-16 :

  - Downloaded brand libtiff-3.7.2.zip, and keep only the libtiff subdir.

  - Make explicit all silent cast from int to uint16 (used VC++7.1, with
    /W4 /WX, and without desabling warning 4244). And all poitner casts
    due to various malloc flavors.

  - Reorder int32/int16 mixed assignements. Should be:
	a32 = b16 = c16 = 16bits expression.
    Not b16 = a32 = c16 = 16bits expression.

  - Operators +=, -=, &= seems to upgrade temporary to 32bits. Operator +/-/& don't
    if both arguments are uint16. Hence we can't use += syntaxe with pointers to 16bits
    data.
    Change :                            To:
    wp[stride] += wp[0]                 wp[stride] = wp[stride] + wp[0]
    wp[0] -= op[0]; wp[0] &= mask;      wp[0] = (uint16)((wp[0] - op[0]) & mask);

  - Added some edits because of Borland bcc32 pickyness.
