Re: [Bug 2136852] [NEW] SB-VM::MOVE-WORD-ARG is not the name of a defined VOP.

There have been a number of recent reports like this one recently, where
the compiler fails with:

"Paul F. Dietz" <[email protected]> writes:

> SB-VM::MOVE-WORD-ARG is not the name of a defined VOP.
>    [Condition of type SIMPLE-ERROR]
>
> Backtrace:
>   0: (SB-C::VOP-PARSE-OR-LOSE SB-VM::MOVE-WORD-ARG)
>   1: (SB-C::GET-OPERAND-INFO #<SB-C:TN-REF :TN #<SB-C:TN t1 :NORMAL> :WRITE-P NIL :VOP SB-VM::MOVE-WORD-ARG>)
>   2: (SB-C::BAD-COERCE-ERROR #<SB-C:TN-REF :TN #<SB-C:TN t1 :NORMAL> :WRITE-P NIL :VOP SB-VM::MOVE-WORD-ARG>)
>   3: (SB-C::EMIT-COERCE-VOP #<SB-C:TN-REF :TN #<SB-C:TN t1 :NORMAL> :WRITE-P NIL :VOP SB-VM::MOVE-WORD-ARG> #<SB-C:TN t2 :ALIAS to X!3> #(NIL NIL T NIL NIL NIL ...) #<SB-C::VOP :INFO SB-C:ALLOCATE-FRAME :A..
> [...]

Thank you for fixing the compiler bugs that this reveals!  However,
there's a second problem, perhaps cosmetic, but I think we should look
at it.  What this is *meant* to do is provide a nice human-readable
internal compiler error about why MOVE-WORD-ARG (or other vops, like
CAR-EQ-IF-LISTP) was not usable in this situation -- some kind of
mismatch between SCs, not being able to find MOVE vops to move between
them.  Instead, we're getting a note that we can't find the
MOVE-WORD-ARG vop by name.  Is this because we have started shaking out
more than we used to?  Judging by versions of SBCL I have lying around,
this change is between 1 and 2 years old, so it's clearly not critical,
but if this is the new normal we should probably change BAD-COERCE-ERROR
not to do things that will always fail.

Christophe