Feature-complete APNG optimization

[ctrlcctrlv]

[fred@デブ.狸.agency ~/.config/OpenToonz/stuff/sandbox/outputs]$ oxipng pcr.apng 
Processing: pcr.apng
APNG files are not (yet) supported

[ctrlcctrlv]

renamed in light of #551; now supports APNG but the support is nearly useless :( [thanks for it anywy tho]

[TPS]

@ctrlcctrlv I think you meant #79 → #511

[AlexTMjugador]

Groundwork has been done to support APNGs (see #511), but the changes were not yet published in a release, so you have to build OxiPNG from source to use them.

Do those changes satisfy your needs, @ctrlcctrlv? If not, why do you find that support "nearly useless"?

[ctrlcctrlv]

fdAT is likewise compressible via the same way* IDAT is compressed so no, probably why @andrews05 wrote they were “not sure if it's something that should actually be added” :-)

* sort of.

[TPS]

@AlexTMjugador Even in #511, @andrews05 says average savings is <1%, so much room for improvement.

I think the idea here is to meet/best apngopt (last release 2015) in a modern fashion like, for static PNG, the original goal was against optipng.

[AlexTMjugador]

Alright, that's fair. I'm renaming this issue to more clearly state what it asks for 😄

[ctrlcctrlv]

exactly. i started thinking about how to do it as well. in theory—

Subsequent frames are encoded in fdAT chunks, which have the same structure as IDAT chunks, except preceded by a sequence number. Information for each frame about placement and rendering is stored in fcTL chunks. The full layout of fdAT and fcTL chunks is described below.

in order to make a well-supported APNG, it's quite important to alternate fcTL and fdAT, as earlier versions of the spec said no chunk may come between an fcTL and its fdAT.

what i think this means is we can use the same code as does the main loop just with &fdat[4:].

likely PngData::from_slice is the primary function that needs changes