[Issue 17347] New: DMD generates different (and wrong) output in -release mode

via Digitalmars-d-bugs digitalmars-d-bugs at puremagic.com
Mon Apr 24 08:57:49 PDT 2017


https://issues.dlang.org/show_bug.cgi?id=17347

          Issue ID: 17347
           Summary: DMD generates different (and wrong) output in -release
                    mode
           Product: D
           Version: D2
          Hardware: x86_64
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P1
         Component: dmd
          Assignee: nobody at puremagic.com
          Reporter: alexander.breckel at gmail.com

The following snippet correctly prints "0" when compiled with dmd (no -release)
or ldc. However, adding -release prints a weird value. The value is slightly
different for each run, but stays in the same number region.

DMD64 D Compiler v2.074.0 (ArchLinux)

$ dmd -release app.d && ./app
140734704773424

I was not able to further reduce the code, as changing seemingly unimportant
things, like removing the "number" field from Wrapper, "fixes" the issue:

import std.stdio;

void main() {
    writeln(f(Wrapper(ABCD.B, 1)).length);
}

enum ABCD { A, B, C, D }

struct Wrapper {
    ABCD abcd;
    uint number;
}

bool[] f(Wrapper x) {
    final switch(x.abcd) {
        case ABCD.A:
        case ABCD.B:
        case ABCD.C:
        case ABCD.D:
            return [];
    }
}



Here's the relevant output of objdump -D, just in case it helps:

000000000043abc8 <_Dmain>:
  43abc8:    55                       push   %rbp
  43abc9:    48 8b ec                 mov    %rsp,%rbp
  43abcc:    48 83 ec 10              sub    $0x10,%rsp
  43abd0:    c7 45 f8 01 00 00 00     movl   $0x1,-0x8(%rbp)
  43abd7:    c7 45 fc 01 00 00 00     movl   $0x1,-0x4(%rbp)
  43abde:    48 8b 7d f8              mov    -0x8(%rbp),%rdi
  43abe2:    e8 0d 00 00 00           callq  43abf4 <_D3app1fFS3app7WrapperZAb>
  43abe7:    48 89 c7                 mov    %rax,%rdi
  43abea:    e8 41 00 00 00           callq  43ac30
<_D3std5stdio14__T7writelnTmZ7writelnFNfmZv>
  43abef:    31 c0                    xor    %eax,%eax
  43abf1:    c9                       leaveq 
  43abf2:    c3                       retq   
    ...

000000000043abf4 <_D3app1fFS3app7WrapperZAb>:
  43abf4:    55                       push   %rbp
  43abf5:    48 8b ec                 mov    %rsp,%rbp
  43abf8:    48 83 ec 10              sub    $0x10,%rsp
  43abfc:    48 89 7d f8              mov    %rdi,-0x8(%rbp)
  43ac00:    48 83 ff 03              cmp    $0x3,%rdi
  43ac04:    77 26                    ja     43ac2c
<_D3app1fFS3app7WrapperZAb+0x38>
  43ac06:    48 8d 05 33 fb 02 00     lea    0x2fb33(%rip),%rax        # 46a740
<_IO_stdin_used+0x10>
  43ac0d:    48 63 0c b8              movslq (%rax,%rdi,4),%rcx
  43ac11:    48 8d 04 01              lea    (%rcx,%rax,1),%rax
  43ac15:    ff e0                    jmpq   *%rax
  43ac17:    31 f6                    xor    %esi,%esi
  43ac19:    48 8d 3d 20 40 24 00     lea    0x244020(%rip),%rdi        #
67ec40 <_D11TypeInfo_Ab6__initZ>
  43ac20:    e8 0f 22 00 00           callq  43ce34 <_d_arrayliteralTX>
  43ac25:    48 89 c2                 mov    %rax,%rdx
  43ac28:    31 c0                    xor    %eax,%eax
  43ac2a:    c9                       leaveq 
  43ac2b:    c3                       retq   
  43ac2c:    c9                       leaveq 
  43ac2d:    c3                       retq   
    ...

--


More information about the Digitalmars-d-bugs mailing list