Re: Bug: loadstring() code validator sometimes incorrectly flags chunks as bad

[Roberto Ierusalimschy <roberto@...>]

> If you have a table constructor of the right size, the integer stored
> in the opcode slot following a SETLIST (C=0) can look like another
> SETLIST (C=0).  The code checker will complain about legitimate jumps
> to the instruction immediately following this integer.

I guess this answers the question "If it's bug fixed into 5.1.4, then
I don't see why 5.2 would need a new pseudo-opcode as the loophole is
already closed."  From the list of recent bugs in the validator, a good
part is due to this trick in SETLIST. We can either fix these bugs one
by one, risking find new ones after some time, or we can fix the cause
of the problem (the existence of a non-instruction in the code), making
the code cleaner and more robust.

BTW, many thanks to all who are sending all these reports. Unlike most
other parts of Lua, the checker is poorly tested. (We would like to
learn how to build a good test script for it.)

-- Roberto