The 2003-08-11-VaListArg UnitTest fails on the Interpreter, on x86 and powerpc.

Do you think this bug is possible to fix, or is there some inherent limitation
in the interpeter?  Do you think it's a good candidate for being a 1.2 bug?

-Chris

I'm not sure -- I haven't successfully analyzed it yet. My only hunch so far is
that I think it only affects programs of the form:

bar (va_list ap) {
 ...stuff involving foo's varargs...
}

foo (arg, ...) {
 va_list ap;
 va_start (arg, ap);
 bar (ap);
 ...
}

Ok, I vaguely remember talking about this at one point.  :)

-Chris

Analysis: The iterator into the array of VarArgs stored in the stack of
ExecutionContexts, which we were using as the Interpreter's va_list, could be
invalidated upon entering a new ExecutionContext (because the stack might need
to be resized, potentially making the VarArgs vector move in memory). 

Fixed:
http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20040223/012213.html
http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20040223/012214.html