• via Digitalmars-d-bugs (36/36) Jun 21 2016 https://issues.dlang.org/show_bug.cgi?id=16189

via Digitalmars-d-bugs <digitalmars-d-bugs puremagic.com> writes:

https://issues.dlang.org/show_bug.cgi?id=16189

          Issue ID: 16189
           Summary: Optimizer bug, with simple test case
           Product: D
           Version: D2
          Hardware: x86_64
                OS: Windows
            Status: NEW
          Severity: critical
          Priority: P1
         Component: dmd
          Assignee: nobody puremagic.com
          Reporter: kkryukov gmail.com

Consider this code:
=== a.d ===
struct T { long a, b; }
T[] data = [{0,0}];
void main()
{
    int a = 0;
    for (int i = 0; i >= 0; i--)
    {
        data[0] = data[a];
        a--;
    }
    assert(a == -1); // Fails with -O
}
=== end ===

Fails when built with "-O", succeeds without "-O". The full command line:
dmd a.d -ofa.exe -O
dmd a.d -ofb.exe

Windows 7, dmd 2.071.0

If you add "-m64" to dmd command, the bug disappears, but re-appears if "int a"
is changed to "long a".

Reduced from a large project with the help of dustmite and a lot of "WTF!".

--