[Issue 12164] New: Function returning ptrdiff_t.min in 64-bit returning 0 when -O is set.

[d-bugmail at puremagic.com]

https://d.puremagic.com/issues/show_bug.cgi?id=12164

           Summary: Function returning ptrdiff_t.min in 64-bit returning 0
                    when -O is set.
           Product: D
           Version: D2
          Platform: All
        OS/Version: All
            Status: NEW
          Keywords: wrong-code
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody at puremagic.com
        ReportedBy: opantm2+dbugs at gmail.com


--- Comment #0 from Kapps <opantm2+dbugs at gmail.com> 2014-02-14 17:58:07 PST ---
The title isn't very descriptive, but perhaps the sample is more useful.
Reduced from std.variant.
Uncommenting the writeln, using -m32, commenting out the if(*rhsPA == *zis)
return 0, or not passing in -O causes the problem to not occur. Otherwise
compile the below with -O -m64 to have it occur.

Sample:

import std.stdio, std.exception, std.c.string;

struct Foo {    

    ubyte[32] store;

    this(A value) {
        memcpy(&store, &value, value.sizeof);
    }

    static ptrdiff_t compare(A* rhsPA, A* zis) {
        if (*rhsPA == *zis)
            return 0;
        //writeln("Returning min");
        return ptrdiff_t.min;
    }

    bool opEquals(Foo rhs) const {
        auto zis = cast(A*)&store;
        auto rhsPA = cast(A*)&rhs.store;
        // Below prints 0 if previous writeln line is commented and -O is set.
        // Otherwise prints -9223372036854775808 as expected.
        writeln(compare(rhsPA, zis));
        return compare(rhsPA, zis) == 0;
    }
}

static struct A { int a; }

void main() {
    enforce(Foo(A(3)) != Foo(A(4)));
}

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------