Stray `One` command leads to `ParseFail(NoCurrentOp)`

[martin-hughes]

This sounds daft, but there's a large number (~40) of uACPI dumps that have a One opcode just floating by itself. Which leads inevitably to ParseFail(NoCurrentOp).

For example, dump 8C21F14AD994 has this snippet in its decompilation:

line 1111 onwards:
    External (TBTF, MethodObj)    // 1 Arguments
    External (XBAS, UnknownObj)

    Name (SS1, 0x00)
    Name (SS2, 0x00)
    Name (SS3, One)
    One // <- Stray
    Name (SS4, One)
    One // <- Stray
    OperationRegion (GNVS, SystemMemory, 0x8FF4F000, 0x0792)
    Field (GNVS, AnyAcc, Lock, Preserve)
    {
(etc.)

(I've added the comments // <- Stray)

Why are those One commands there? Who knows... iasl won't even compile a snippet like that! But it crashes the parser.

Sadly being unable to compile this in iasl makes it hard to test. Might need a bit more test infrastructure 🤓

[IsaacWoods]

Yes - I've come across these (you also see random Zeros appear in places) and don't really know where they come from. My best guess is some kind of AML patching tool that (accidentally, one assumes) leaves 0x00 or 0x01 bytes in places they shouldn't be.

I think we should be able to handle this surprisingly elegantly by gracefully handling trying to contribute an argument when we don't have a current in-flight op. We actually already have a help that does this (MethodContext::contribute_arg) but are not currently using it at the top-level. Whether we want to warn when we throw away an argument - unknown.

I'll open a PR with a potential fix. I don't currently have a copy of the uACPI dumps around to test so would be very grateful if you could run it over that PR (#293)?

[martin-hughes]

👍 cheers Isaac. All looks good to me, and seems to remove the issue from the uACPI dumps.

Separately, just for you to think about - I was wondering if "ignore extraneous arguments" is quite right. But I also can't think of anything better. For example, emitting a warning for things like this stray opcode could be a nuisance for the end-user.