The bytecode reader seems to fail on the attached input.  I suspect that it is a
bug in the verifier, since the problem occurs from a switch statement that
selects on a uint with each possible case typed as an int

This was reported by Mark Hills:

llvm-as file.ll
opt file.bc -o null.bc
opt: /home/vadve/criswell/llvm/include/Support/Casting.h:197: typename
llvm::cast_retty<To, From>::ret_type llvm::cast(const Y&) [with X =
llvm::Constant, Y = llvm::Value*]: Assertion `isa<X>(Val) && "cast<Ty>()
argument of incompatible type!"' failed.
opt[0x861e78a]
opt[0x861e9aa]
opt[0x42029188]
opt(abort+0x16e)[0x4202a8c2]
opt[0x42022ecb]
opt(llvm::cast_retty<llvm::Constant, llvm::Value*>::ret_type
llvm::cast<llvm::Constant, llvm::Value*>(llvm::Value* const&)+0x3b)[0x83bcbd7]
opt(llvm::BytecodeParser::ParseInstruction(unsigned char const*&, unsigned char
const*, std::vector<unsigned, std::allocator<unsigned> >&,
llvm::BasicBlock*)+0x11f4)[0x83af6ec]
opt(llvm::BytecodeParser::ParseInstructionList(llvm::Function*, unsigned char
const*&, unsigned char const*)+0x2ca)[0x83b2ffc]
opt(llvm::BytecodeParser::materializeFunction(llvm::Function*)+0x58a)[0x83b3ae6]
opt(llvm::BytecodeParser::materializeModule()+0x5c)[0x83c2b76]
opt(llvm::ModuleProvider::releaseModule()+0x18)[0x83c8238]
opt(llvm::BytecodeParser::releaseModule()+0x11)[0x83c2b91]
opt(llvm::ParseBytecodeFile(std::basic_string<char, std::char_traits<char>,
std::allocator<char> > const&, std::basic_string<char, std::char_traits<char>,
std::allocator<char> >*)+0x39)[0x83b7b8d]
opt(main+0x6b)[0x83a5efb]
opt(__libc_start_main+0x95)[0x42017589]
opt(std::basic_ostream<char, std::char_traits<char> >& std::operator<<
<std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&,
char)+0x5d)[0x83a5de1]
Abort (core dumped)

Created an attachment (id=118) [details]
Code to reproduce the problem

This is code from Mark that reproduces the problem.

Created an attachment (id=119) [details]
Reduced test case

This is a reduced test case.  It is the same as the previous test case, but
isolates the problem down into the switch statement that Mark suspected was the
problem.

Interesting.  Does the problem go away when you change the int constant to be a 
uint?  This is probably a matter of having the verifier assert that the 
constatns have the same type as the switched value... i.e. that this is 
malformed LLVM code.

-Chris

Yes, the bug goes away when the switch selector and labels are all uints.

Okay, then this is a verifier bug (for not catching it) and the input is 
malformed.  I'll make patch for the verifier.

-Chris

Fixed.  Testcase here:
test/Regression/Verifier/2004-05-21-SwitchConstantMismatch.ll 

Patch here:
http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20040517/014436.html

-Chris