[llvm-bugs] [Bug 35750] New: __builtin_mul_overflow with mixed sign int operands incorrect result

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=35750

            Bug ID: 35750
           Summary: __builtin_mul_overflow with mixed sign int operands
                    incorrect result
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: release blocker
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: ruttenberg at reservoir.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 19603
  --> https://bugs.llvm.org/attachment.cgi?id=19603&action=edit
Reproducer -- passes with gcc and clang 4.0

Attaching a test simplified from the gcc test suite.

Essentially unsigned int = ((signed) -1) * 1 == 1

Gcc gets ((unsigned) -1) as the result, which I think is right.

The bug originated with this change:

Author: Vedant Kumar <vsk at apple.com>
Date:   Sat Dec 16 01:28:25 2017 +0000

Previously, the signed operand was sign extended to 33 bits and the unsigned
operand was zero extended to 33 bits and the multiply was carried out in
extended precision, which seems right.

Now we just negate the first operand if it is negative, which seems wrong.

-- 
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/20171226/a97c15b7/attachment.html>