Possible edge case in decoder when very first OP is QOI_OP_RUN

[phoboslab]

There's an edge-case that currently is handled differently by various decoders. At a first glance it looks like this decoder may not be spec-conformant. The spec is not clearly worded in this case, but I have decided to interpret it in a particular way now. Sorry about that!

Please see the original issue here phoboslab/qoi#258 for the details.

[AregevDev]

Hi, this shouldn't be hard to fix. Just add the starting pixel color to the index array.

I have a a question for @phoboslab though:
Does the spec allow the encoder to start with a run operation? You mentioned you never do that in the reference encoder, but other encoder implementations may do.

@aldanor If that's an acceptable solution, I can open a PR in the coming days. Also, I have a few specific questions about the implementation of the encoder / decoder. How can I reach out?

[aldanor]

Yes; should add the edgecase.qoi to the current test suite and see what happens. Should be an easy fix if it's misinterpreted.

[AregevDev]

I wrote a test that reads edgecase.qoi and decodes it and works as intended.

I still need to debug the decoding process though, I am not sure when it puts the initial opaque black pixel in the seen array.