[Pnet-developers] Problems in ternary with upcasting

[Gopal V]

Hi,
        This bug is either something MS forgot to tell ECMA or 
somebody didn't think verifiers were that important :)

ECMA 11.4 -- Control Flow , line number 39 

 4. Each slot on the stack shall have the same data type at any given point 
    within the method body, regardless of the control flow that allows 
    execution to arrive there.

So consider this code sequence

        brfalse ?L1
        ldloc.1 // which is class 'A'
        br ?L2
?L1:
        ldloc.2 // which is class 'B'
?L2:
        
In fact this is actually code pumped out by the MS compiler (as well as
our CSCC) for

aORb ? (Object) (a) : (Object) (b);

The entire issue revolves round the fact that the code is valid , but
not verifiable !! (at least not according to the ECMA spec :)

The issue could have been resolved by explicit "castclass" dumped in
the IL code , but "somebody had blundered" ... thankfully it's not me.

Any suggestions anyone ?
Gopal
-- 
The difference between insanity and genius is measured by success

ternary.cs
Description: Text document