Backward compatibility: -DAVIF_CODEC_RAV1E=OFF -DAVIF_LOCAL_RAV1E=OFF results in a warning about AVIF_LOCAL_RAV1E not used by the project about libavif HOT 6 CLOSED

The AVIF_LOCAL_RAV1E variable is indeed unused when AVIF_CODEC_RAV1E is OFF, so the unused AVIF_LOCAL_RAV1E warning makes sense to me.

I suggest leaving it as is but if the warning is too annoying, according to this thread one can either use --no-warn-unused-cli or remove -DAVIF_LOCAL_RAV1E=OFF.
We can also add set(ignoreMe "${AVIF_LOCAL_RAV1E}") but it does not look as good practice.

from libavif.

My question is whether we intend to be backward compatible. I don't get the cmake warning if I run that cmake command with libavif 1.0.2, for example.

There is a subtle difference between "not used" and "used but has no effects". In libavif 1.0.2, -DAVIF_LOCAL_RAV1E=OFF is used but has no effects when -DAVIF_CODEC_RAV1E=OFF is specified.

from libavif.

This is similar to passing -DAVIF_ENABLE_GTEST=OFF -DAVIF_LOCAL_GTEST=OFF to cmake on the main branch. There is no warning about AVIF_LOCAL_GTEST being not used.

from libavif.

I don't get the cmake warning if I run that cmake command with libavif 1.0.2, for example.

To me, the presence or absence of a warning has no impact on the backward compatibility, as long as it works.

This is similar to passing -DAVIF_ENABLE_GTEST=OFF -DAVIF_LOCAL_GTEST=OFF to cmake on the main branch. There is no warning about AVIF_LOCAL_GTEST being not used.

So this is rather a consistency question than a backward compatibility question?
If so, we can use set_local_or_system_option for all fields, not only libyuv and libsharpyuv.

We talked about that with @vrabaud earlier today and the conclusion was that there was little interest in adding the capacity to completely disable zlib/png dep (off/local/installed choice instead of the current local/installed). However, having -DAVIF_LIBYUV=LOCAL next to -DAVIF_LOCAL_ZLIBPNG=ON is inconsistent, so it may be a sufficient reason to use set_local_or_system_option everywhere, on top of what you mentioned.

from libavif.

What I can do is display a message that this is the old way of doing things and what the new way is. I'll check whether the variable exists, is ON or OFF and that will also remove the warning.

from libavif.

I understand Vincent wanted to support the old AVIF_LOCAL_ variables. That's why I reported this warning message. The current support for AVIF_LOCAL_RAV1E is good enough, but it doesn't fully match the original support.

Vincent: The deprecation message you suggested is a good idea.

from libavif.