[LLVMbugs] [Bug 23212] New: Regression: Miscompile with -O1 and -fsanitize=signed-integer-overflow

[bugzilla-daemon at llvm.org]

https://llvm.org/bugs/show_bug.cgi?id=23212

            Bug ID: 23212
           Summary: Regression: Miscompile with -O1 and
                    -fsanitize=signed-integer-overflow
           Product: clang
           Version: trunk
          Hardware: Macintosh
                OS: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: jonathan.sauer at gmx.de
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

The following program should print 1, and it does so when compiled with clang
r234732 and either -O1 or -fsanitize=signed-integer-overflow. However, when
compiled with both -O1 *and* -fsanitize=signed-integer-overflow, it prints 0.

#include <cstdio>

int flipSign(unsigned int t)
{
  return -int(t);
}

int main()
{
  std::printf("%d\n", -2 == flipSign(2));
}


% ~/LLVM/build/Release+Asserts/bin/clang++ -O1 clang.cpp && ./a.out
1

% ~/LLVM/build/Release+Asserts/bin/clang++ -fsanitize=signed-integer-overflow
clang.cpp && ./a.out
1

% ~/LLVM/build/Release+Asserts/bin/clang++ -O1
-fsanitize=signed-integer-overflow clang.cpp && ./a.out
0

When compiled with clang r234147, it always prints 1. So does removing the
explicit cast to "int" in "flipSign".

The miscompile also happens with higher optimization levels, e.g. -O2, -O3 and
-Os.


Detailed clang version:
clang version 3.7.0 (trunk 234732)
Target: x86_64-apple-darwin13.4.0
Thread model: posix

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20150413/39e37fd8/attachment.html>