[MC, RISCV] Fixed StringRef Assertion `Index < Length && "Invalid index!"'

[apazos]

Summary

Handle the case IDVal is an empty string.

This bug was uncovered by a LLVM MC Assembler Protocol Buffer Fuzzer for the RISC-V assembly language.

[asb]

test/Assembler/empty-string.s

1 ↗

(On Diff #160914)

This test shouldn't actually require asserts should it?

[apazos]

Thanks Alex for reviewing. Adding some more reviewers who might +1 before I merge

[rnk]

lgtm

lib/MC/MCParser/AsmParser.cpp

1845 ↗

(On Diff #162516)

I'd probably spell it IDVal.startswith(".") for readability.

[apazos]

lib/MC/MCParser/AsmParser.cpp

1845 ↗

(On Diff #162516)

good suggestion, will update the patch and merge. Thanks, Reid.

llvm/trunk/lib/MC/MCParser/AsmParser.cpp

bool AsmParser::parseStatement(ParseStatementInfo &Info,
   if (areMacrosEnabled())
     if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) {
       return handleMacroEntry(M, IDLoc);
     }
 
   // Otherwise, we have a normal instruction or directive.
 
   // Directives start with "."
-  if (IDVal[0] == '.' && IDVal != ".") {
+  if (IDVal.startswith(".") && IDVal != ".") {
     // There are several entities interested in parsing directives:
     //
     // 1. The target-specific assembly parser. Some directives are target
     //    specific or may potentially behave differently on certain targets.
     // 2. Asm parser extensions. For example, platform-specific parsers
     //    (like the ELF parser) register themselves as extensions.
     // 3. The generic directive parser implemented by this class. These are
     //    all the directives that behave in a target and platform independent

llvm/trunk/test/Assembler/empty-string.s

+// REQUIRES: riscv-registered-target
+// RUN: not llvm-mc -triple riscv32-unknown-linux-gnu < %s 2>&1 | FileCheck %s
+"" # CHECK: error: unrecognized instruction mnemonic