Problem with lzfse_decode_buffer_with_scratch about lzfse HOT 1 OPEN

Is this also related to #44 ?

TLDR: It's related but does not solve the issue because there's an unhandled corner case.

I withdrew that commit as there is an unhandled corner case (*). Sorry.

I'm not sure I have any good answers for you but here are my two cents. Most of which I suspect you already know.

Within the context of lzfse_decode_buffer my understanding is that the return value, let's call it ret, can communicate three outcomes:

1. Success. Here ret reflects the number of bytes written into the destination buffer.
2. Destination buffer is too small. Here ret equals the size of the destination buffer.
3. Decode error. Here ret equals 0. This is not explicitly documented but we can see the relevant code here.

The issue is that ret is overloaded. It's not always possible to assign an outcome based on ret alone.

For example, let's take a ret value of zero. This could mean:

1. Success. Where we've decoded a zero-length input file.
2. Destination buffer is too small. Where we've passed a zero-sized destination buffer.
3. Decode error. Where we've passed invalid data.

Although we can exclude scenario 2 by not passing in zero-sized buffers, we still cannot differentiate between scenarios 1 and 3 (*). We need more information. At some level, we need to tweak the API. One such suggestion is here.

from lzfse.