From: John B. <joh...@gm...> - 2009-09-28 18:03:36
|
From: Glenn Randers-Pehrson [mailto:gl...@gm...] >Width is read by png_get_uint_31(), so the failure is probably there. >It's probably as I originally suspected a BIG_ENDIAN vs LITTLE_ENDIAN >problem. For that to be correct PNG_BIG_ENDIAN_SUPPORTED would have to have been defined - since the code in absence of this #define works correctly regardless of the machine architecture. (None of the standard code in libpng needs to know the byte sex of the hardware does it?) The bug would also be reproducible on 1.4.0beta82, assuming everything has been compiled with the same options, but: From: Arnaud BOEGLIN [mailto:a.b...@gm...] >I tried the version 1.4.0beta82 and it works ! Hum. What is the width that is going wrong? I.e. exactly what value is failing the test? If we know that we will be able to make far more informed guesses. John Bowler <jb...@ac...> |