Update free_functions.rs

[neckaros]

Support guessing AVIF image generated by this library (with native avif)

Exemple of image generated by this library with the magic bytes not yet recognized:
image.zip

[fintelia]

The array size needs to be updated, but otherwise looks good to me

[neckaros]

The array size needs to be updated, but otherwise looks good to me

You are right my bad I will update it

[neckaros]

I ran the test locally and the failing test do not seem related to my commit but I could be wrong?

[fintelia]

Thanks!

[kornelski]

These bytes before ftyp are the size of the ftyp chunk. This chunk contains a variable-length list, so it could have any size.

Instead of collecting implementation-specific lengths, can the length be allowed to be any bytes? The two most significant bytes could be required to be 0, since the chunk is unlikely to ever exceed 64K.

[neckaros]

These bytes before ftyp are the size of the ftyp chunk. This chunk contains a variable-length list, so it could have any size.

Instead of collecting implementation-specific lengths, can the length be allowed to be any bytes? The two most significant bytes could be required to be 0, since the chunk is unlikely to ever exceed 64K.

I wondered it myself but could not see how to do it with the current implementation

[kornelski]

A simple solution could be to add a start offset to the table of magic strings, and skip 4 bytes in AVIFs case.

[fintelia]

If we included a mask of bytes to ignore, we could also have an actual fix for #660