[llvm-bugs] [Bug 27848] New: Incorrect optimization in memcpyopt with longjmp and volatile locals

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=27848

            Bug ID: 27848
           Summary: Incorrect optimization in memcpyopt with longjmp and
                    volatile locals
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: eli.friedman at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Testcase:

#include <setjmp.h>
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
int zz = 1;
__attribute((noinline))
void g(volatile int* y, jmp_buf *b) {
  *y = 10;
  if (zz) longjmp(*b, 2);
}

__attribute((noinline))
int f() {
  volatile int x, y;
  x = 1;
  jmp_buf b;
  if (setjmp(b)) return x;
  g(&y, &b);
  memcpy((void*)&x, (void*)&y, 4);
  return x;
}

int main() { if (f() != 1) abort(); }

Runs correctly with gcc or clang -O0; aborts with clang -O2.  I'm pretty sure
it's a bug in memcpyopt: it's trying to do call slot optimization on y.

-- 
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/20160524/41c1069b/attachment-0001.html>