libjpeg allows specifying the marker interval either in MCU blocks or in
MCU rows. Support both, via separate parameters, rather than requiring
callers to do the math.
Co-authored-by: Andrew Murray <radarhere@users.noreply.github.com>
tp_print was renamed to tp_vectorcall_offset with Python 3.8, and tp_compare was renamed to tp_as_async with Python 3.5.
tp_size has always been tp_basicsize; I don't know why that one was wrong.
Getters are supposed to have signature "PyObject *(PyObject *self, void *closure)", but the closure argument is often not used.
In wasm it causes a trap if a function is declared with one argument and then called with two.
The file `libImaging/Imaging.h` gets installed flat on
Ubuntus into the public Python includes.
When building a newer version of Pillow from source and
hinting the Python includes "too early" in includes, e.g.
in package managers, this can confuse the two files and
pick up the external file over the internal one. With
different versions, this mismatch can lead to build errors,
e.g. undefined macros.
The most robust way to avoid this is to pre-fix the internal
include accordingly, so that the relative path to the including
file has to match as well.
* Improve encoding of TIFF tags
- Pass tagtype from v2 directory to libtiff encoder, instead of
autodetecting type.
- Use explicit types. E.g. uint32_t for TIFF_LONG to fix issues on
platforms with 64bit longs.
- Add support for multiple values (arrays). Requires type in v2
directory and values must be passed as a tuple.
- Add support for signed types (e.g. TIFFTypes.TIFF_SIGNED_SHORT).
Co-authored-by: Andrew Murray <radarhere@users.noreply.github.com>
Uses JPEGQUALITY pseudo-tag from libtiff.
Also changes the way tags are passed to PyImaging_LibTiffEncoderNew from
dict to list to ensure that COMPRESSION tag is added before JPEGQUALITY.
This is required as the COMPRESSION tag registers the JPEGQUALITY
pseudo-tag.
- Pass tagtype from v2 directory to libtiff encoder, instead of
autodetecting type.
- Use explicit types. E.g. uint32_t for TIFF_LONG to fix issues on
platforms with 64bit longs.
- Add support for multiple values (arrays). Requires type in v2
directory and values must be passed as a tuple.
- Add support for signed types (e.g. TIFFTypes.TIFF_SIGNED_SHORT).
