* Support for parsing boxes: a1op, a1lx, lsel
* Added operating point and "skip" support to avifCodec impls for layer selection support
* Reworked avifdec decoding to support nth image decoding and minor optimization
* New avifProgressiveState enum to expose info about progressive AVIFs to users of libavif
* New avifdec args: --progressive, --index

(rough draft, scale.c needs more work)

…tching

…able

Add diagnostic messages for all AVIF_RESULT_DECODE_ALPHA_FAILED or
AVIF_RESULT_DECODE_COLOR_FAILED errors.

* Make avifDecoderItemMaxExtent progressive-aware

Make libavif speed setting to map 1-to-1 to libaom speed setting.
Adjust usage for speed >= 7 according to AOM_USAGE_ALL_INTRA
availability and sequence coding.

…tly (#687)

* Fix avifDecoderItemMaxExtent() - compensate for sample->offset correctly

Ideally should use this line and PROJECT_VERSION_* variables to generate include/avif/avif.h version info to keep things nicely in sync

Clarify the comment for the spatialID member of the avifDecodeSample
struct. Avoid the ambiguous word "otherwise".

Fix #697.

In avifCodecDecodeInputFillFromDecoderItem(), the assertion

        assert(lselProp->u.lsel.layerID < MAX_AV1_LAYER_COUNT);

is apparently intended to ensure that the following uint8_t cast is
safe:

        sample->spatialID = (uint8_t)lselProp->u.lsel.layerID;

So move the assertion next to that statement.

Call avifDecoderNextImage() in an assignment statement in the while
loop's conditional expression. This allows us to have only one
avifDecoderNextImage() call.

Improve the error message for an unsupported output file extension. If
we reach here, we recognize the file extension (currently only .avif),
but the file cannot be the output from avifdec. So the error message
"Unrecognized file extension" is a little confusing.

Remove the unneeded null check for 'decoder' under the 'cleanup' label.
The 'decoder' variable is initialized as follows:

    avifDecoder * decoder = avifDecoderCreate();
    decoder->maxThreads = jobs;
    decoder->codecChoice = codecChoice;
    ...

so it cannot possibly be a null pointer.

Compare partialByteCount with item->size - offset rather than
item->size, because the read should start from 'offset'.

The change "Don't need to disable MSVC warnings 5031 and 5032 (#681)"
is a change to commit 5a98790 (the
"-j all" option of avifenc and avifdec), which was added during the
0.9.2 development cycle. So this change is not a change to older
releases and does not need to be documented as a change in 0.9.2.

Avoid the heap allocation of aomImage plane buffers and copying of input
image in aomCodecEncodeImage(). This requires manually setting up the
aom_image_t struct 'aomImage'.

Fix #694.

…AnyToRGBAnySlow()

The old comment describes the behavior when this bit is disabled. The
comment should describe the behavior when this bit is enabled.

Fix #715.

Use full paths and move configure w/install

Also fix a comment typo (libaom, not libavif).

Fixes: #706

This avoids the following dav1d_log() message in 32-bit builds:
    Frame size limit reduced from 268435456 to 67108864.

Commit ea3e97e eliminated the
unnecessary copying of input image data into the aom_image_t struct, but
it exposed a libaom bug (https://crbug.com/aomedia/3113). Work around
the libaom bug by resurrecting the original code that copies input image
data and using it when the image width or height is equal to 1.

Skip the YUV<->RGB conversion tests except for the first loop iteration
(loop == 0) and the first image of an image sequence. The YUV<->RGB
conversion tests are time-consuming for large images, so we should try
to avoid duplicate tests.

Adapted from Joe Drago's pull request
#527, with the limitation
that decoder->imageSizeLimit must be less than or equal to the default
value and must not be set to 0 (reserved for future use). This way we
don't need to audit our code for integer overflows due to a large image
width or height.

Set decoder->imageSizeLimit to 11 * 1024 * 10 * 1024 in
avif_decode_fuzzer.cc to keep its memory consumption under 2560 MB.

#263

Since avifImageRGBToYUV() ignores rgb.chromaUpsampling, we only need to
test avifImageRGBToYUV() with a single upsamplingsIndex.

Add custom diagnostic messages if item size, track size, or grid
dimensions exceed imageSizeLimit.

Related to #263.

* Check for scale values that cause overflow

This addresses #734

…he buffer the item's full size

This avoids progressive image partial item reads causing reallocs of this buffer, which can cause
dangling pointers in the underlying AV1 decoders (use-after-free). As the natural pattern of this
type of usage is to ultimately decode the final image, those additional reallocations are also
typically unnecessary overhead, as we'll eventually be using this sized buffer anyway.

This should address Chromium issue 1239472.

…d incorrect alphaRowBytes math

Allow progressive decoding and use a non-persistent avifIO object. This
allows avif_decode_fuzzer to reproduce the AddressSanitizer
heap-use-after-free error in https://crbug.com/1239472.

…, infer ispe from color item

See: https://crbug.com/1245673

…property, infer ispe from color item"

This reverts commit efed118. That
commit is intended for Chrome 93 only.

See https://code.videolan.org/videolan/dav1d/-/commit/753eef833b:
Merge the 3 threading models into a single one
by Victorien Le Couviour--Tuffet <victorien@videolan.org>

Improve the check added in
#735.

Compute maxFixedValueWidth and maxFixedValueHeight in 64-bit
multiplication by casting the first operand to int64_t. Otherwise, the
first operand is an int and the second operand is a uint32_t, so the
multiplication is 32-bit.

Replace 0x7fffffff with the symbolic constant INT_MAX because the
product will be stored in an int variable.