[llvm-bugs] [Bug 33654] New: x86 assembler immediates should be limited to 32 bit

via llvm-bugs llvm-bugs at lists.llvm.org
Fri Jun 30 05:25:15 PDT 2017


https://bugs.llvm.org/show_bug.cgi?id=33654

            Bug ID: 33654
           Summary: x86 assembler immediates should be limited to 32 bit
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: dimitry at andric.com
                CC: llvm-bugs at lists.llvm.org

In bug 33386, Jonas Vekeman encountered a situation where OpenSSL's inline
assembly in combination with function inlining could sometimes lead to errors
from the inline asm parser, e.g.:

    <inline asm>:1:12: error: invalid operand for instruction
            divq      $-8446744073709551616
                      ^~~~~~~~~~~~~~~~~~~~~
    LLVM ERROR: Error parsing inline asm

For this particular case, I created a test.s file with just:

    divq 0x7fffffff
    divq 0x80000000
    divq 0xffffffff
    divq 0x100000000
    divq 0xffffffffffffffff

For this assembly, GNU as rightfully complains about lines 2 through 4:

    divtest.s: Assembler messages:
    divtest.s:2: Error: operand type mismatch for `div'
    divtest.s:3: Error: operand type mismatch for `div'
    divtest.s:4: Error: operand type mismatch for `div'

However, clang 5.0.0 doesn't complain at all, and produces an object file.
Disassembling it shows that clang has silently wrapped the too-large values:

       0:       48 f7 34 25 ff ff ff 7f         divq    2147483647
       8:       48 f7 34 25 00 00 00 80         divq    -2147483648
      10:       48 f7 34 25 ff ff ff ff         divq    -1
      18:       48 f7 34 25 00 00 00 00         divq    0
      20:       48 f7 34 25 ff ff ff ff         divq    -1

At the very least, it should complain here like GNU as does, IMHO.

-- 
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/20170630/08558811/attachment.html>


More information about the llvm-bugs mailing list