from the new call to png_read_gray_to_rgb() (bug introduced in
libpng-1.4.0beta67). (Patch contributed by Alam Arias)

to run faster in the usual PNG_ALL_FILTERS case.  The previous, more
accurate test, would save some computation in uncommon cases, but is
probably not worth while in the long run.

after the 'png_debug(1, "in png_write_find_filter");' statement.

    to obtain the same result as previously when the user asks for
    a filter selection that does not include NONE or SUB, e.g., if
    the user asks for PAETH-only that's what they should get.

to obtain the same result as previously when the user asks for
a filter selection that does not include NONE or SUB, e.g., if
the user asks for PAETH-only that's what they should get.

The AVG and PAETH filters will never be selected for the first
row, so we need not test them.

contributed by Alam Arias

and png_write_png().

to see if that works better for the auto-license-detectors

outside of the PNG_READ|WRITE_ANCILLARY_CHUNK_SUPPORTED blocks, where
they did not really belong.  Updated CHANGES and ANNOUNCE.

It is not of any use in libpng-1.4.0; it was only to support binary
compatibility between some old libpng-1.0.x versions.  It should have
been removed from libpng-1.2.x as well but it is too late for that now.

    to "#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED", to make the png_set_keep
    mechanism available for ignoring known chunks even when not saving
    unknown chunks.

about calling png_set_IHDR() multiple times and about the order
of color samples in the tRNS chunk.

about calling png_set_IHDR() multiple times and about the order
of color samples in the tRNS chunk.

about using png_set_io_ptr() in configure scripts for detecting libpng