Also omit the 'struct' keyword before avifDiagnostics.

Using 'const' on non-pointer function parameters means the function
cannot modify the parameters. This could be useful in function
implementations, but should be avoided in function declarations.

C++ allows the function declaration and definition to differ in the
constness of the parameters. For example, in add.h we can declare

    int Add(int a, int b);

and in add.cc we can define the function with this prototpye:

    int Add(const int a, const int b)
    {
        return a + b;
    }

Unfortunately C doesn't allow this, even though gcc and clang allow this
as a language extension.

Currently one ends up with the absolute path on the build machine.

#637)

* Merge avifCodec's open call with its getNextImage call to avoid codec init during parse, and simplify the codec API

Fixes: #636

…ne), remove extraneous newline

The diagnostic messages in avifParseItemPropertyContainerBox() should
say "Box[ipco]", not "Box[iprp]".

Also remove two unnecessary int32_t casts. cropY.n and cropY.d are
already of the int32_t type.

Simplify the conditionals in avifCodecConfigurationBoxGetFormat(). Once
monochrome has been eliminated, YUV 4:2:0 is the only one with
subsampling_y == 1. And the remaining two can be differentiated by
testing subsampling_x.

Remove the avifDiagnosticsClearError() call from
avifEncoderSetCodecSpecificOption() because
avifEncoderSetCodecSpecificOption() returns void and doesn't fail.

Change "decode error" to "decode or encode error" in the description of
avifDiagnostics.

Capitalize "switch".

P => Q is equivalent to !P || Q. So it is not necessary to test P again
in !P || (P && Q).

If a function has an avifDecoderData * data parameter, it can use
data->diag. It is not necessary to add an avifDiagnostics * diag
parameter to the function.

The box name of avifParseChunkOffsetBox depends on largeOffsets.

In calcCenter(), set the numerator to dim directly if dim is odd, rather
than calculate it from dim >> 1.

In clapFractionCD(), use *= whenever possible.

Also remove extraneous parentheses.

Fix #657.

Avoid the widthN * 2 and heightN * 2 multiplications, which could
overflow int32_t, by performing the widthN / widthD and
heightN / heightD divisions early. Those two divisions are needed to
calculate cropRect->width and cropRect->height.

…eplaces 'axis' with 'mode' and inverts the behavior

In the AV1 bitstream, chroma_sample_position is assigned a value if both
subsampling_x and subsamplingy are equal to 1. This actually includes
both monochrome and YUV 4:2:0, but chroma_sample_position is not used in
monochrome, so we only print chroma_sample_position for YUV 4:2:0.

Check if a uint32_t variable is <= INT32_MAX before casting it to
int32_t.

Perform the arithmetic operations in int64_t and then check for int32_t
overflows.

Change clapFractionCD(), clapFractionAdd(), and clapFractionSub() to
return avifBool. They return AVIF_FALSE if any arithmetic operation
overflows int32_t.

Fix the clang -Wunused-macros (macro is not used) warning about the
HAVE_AOM_CODEC_SET_OPTION macro when libavif is configured to use libaom
for decode only (-DAVIF_CODEC_AOM_ENCODE=OFF
-DAVIF_CODEC_AOM_DECODE=ON).

1. Have avifCropRectConvertCleanApertureBox() and
avifCleanApertureBoxConvertCropRect() call avifDiagnosticsClearError()
on entry.

2. avifDiagnosticsClearError() just needs to set the first character of
the diag->error buffer to '\0' rather than memset the entire buffer.

Updates the libaom version checked out in the ext and docker scripts
from v3.1.0 to v3.1.1.

We don't need to allocate codec->internal->svt_config from the heap.

Add extern "C" blocks to tests/compare.h and tests/testcase.h to allow
tests written in C++ to include these two headers.

…s on the machine

Fixes: #645

Since #pragma warning(push) and #pragma warning(pop) are not used in
this file, we cannot possibly get MSVC warnings 5031 and 5032. Therefore
it is not necessary to disable these two warnings.

Fix #679.