Wunk

I replicated similar crashes going on that I'll investigate soon. It seems at some point on the much larger images it crashes from getting an RLE stream size-prefix with an...

Here's what I gathered from reverse engineering that part of the code again, which my code pretty much mirrors. It looks like the RLE streams are being interpreted right but...

The small one looks like a good test-subject. I'll walk through the file by-hand and see whats up with these RLE streams.

Sai is strictly RLE from what I've seen. The method it uses to write and read the pixel data is all RLE or RLE with a strided offset to fill...

> * ..._does not crash_. ... **File Size is 0.980 MiB** > * ..._segment faults_ ... **File Size is 23.5 MiB**. > * ..._segment faults_ ... **File Size is 23.8...

Did some research and found that the `Flags` field of the table-pages actually are block indexes of where the "next chunk" of a file is located like a singly linked...

Still trying to fix this issue. WHile I have updated the code into the "linked list" nature of the Table-Pages I'm still running into issues and I'm trying to hand-decipher...

Still looking into this. I've reverse engineered Sai's own reading routine and I'll probably just aim to mimic the patterns seen here in my own code. Except for the part...

Thanks for the reference files! Once libsai reaches MVP I def want to look into reverse engineering the sai2 format if there is certainly enough demand/use for such a thing.

`16447` has a bit pattern `0100000000111111` which might imply some kind of alternate encoding is going on such as fixed point or some other type of packed binary format or...