Going to have to grab some other patches for the older era source code off of my main laptop's Windows install due to them being hardcoded to the original developer's file path, lol.

Since I haven't found where in the early source where the elusive bug is yet, using the early source is needed.

Found them! As I thought, they were on my main laptop's Windows 10 install.

Those are being copied to the Microsoft-Stuff share now.

...and now I will have to track down my Windows 10 ISOs.

Thought I added them to the 'Microsoft-Stuff' share some time ago, but they are not present. Windows 11 is there, but without certain workarounds, they are useless to the NUC I am going to use.

Oh, not grabbing Windows 11 24H2, either.

It actually looks to be a bug/oversight either introduced very early on or from the start.

SSEQPlayer doesn't have the issue.

...

This one will be quite tricky to resolve.

Fun fact: that former gaming install (didn't mention that it was tied to an Intel NUC in a later life, hence it being UEFI'd) had a major mishap at one point...

cwcyrix.nsupdate.info/gnu-social/p...

Of course, that requires imaging the other old Windows 10 install (this one was a former gaming install that was Windows 7 -> Windows 10 then UEFI'd...) to the NAS.

There's other data on the main web server that's ~803 GB in size that needs to go on the NAS as well.

So, not the buffer, not the ADPCM code.

I didn't analyze the SWAVs for BW/B2W2 yet; not sure that may yield a clue, however.

Found out that kode54's foobar2000 2sf decoder 0.26 has issues, so 0.25.5 was to be used instead.

Also looks like converting some 2sfs to NCSFs may introduce issues anyway, but I'm not exactly at a Windows setup to test that theory, either.

If so, this goes in the affected list when I create changelogs whenever I release an NCSF rip. (For portamento oversights, there are ~4 known for BW, probably ~10 for B2W2.)

Sonic Rush is the only released example for now due to the historical PSG bug.

Should say 'in some way'.

Some weeks ago, I extracted other samples in Pokémon White 2, and I should see if some of those files are ADPCM. This will apply to its prequel.

Though, that means I would have to compile even older builds to test this particular theory... vs. the ring buffer, which may be looking less likely now.

Apparently this doesn't happen on MacOS 14, so it is likely a regression.

...which wouldn't be surprising to me, either.

Looking less like a Rust bug but more of a MacOS bug, which would not be surprising to me.

...Someone should try this on older MacOS builds if possible.

Uneventful success. Did have to re-enable XMP on the memory, but that was all.

On a hunch, I started reflowing a few SMD components in the following order:

CPU
Work RAM
VRAM
Z80
Z80 RAM
Main ASIC

...it was the Main ASIC; it had unnoticeable solder cracks to the PCB. The Model 2 Genesis came back to life! It still works to this day.