Given the program below ...

[test.d]
<code>
debug(1) import std.stdio;
const int anything = -1000; // Line #2
dchar[] SomeFunc( dchar[] pText, out int pStopPosn)
{
    if (pText.length == 0)
        pStopPosn = 0;
    else
        pStopPosn = -1;
    debug(1) writefln("DEBUG: using '%s' we get %d", pText, pStopPosn);
    return pText.dup;
}

int main(char[][] pArgs)
{
    int sp;

    SomeFunc("123", sp);
    debug(1) writefln("DEBUG: got %d", sp);
    assert(sp == -1);

    SomeFunc("", sp);
//    if (sp != 0){} // Line #22
    debug(1) writefln("DEBUG: got %d", sp);
    assert(sp == -1);
    return 0;
}
</code>

and compiled with "dmd test.d" version 0.125, the last assert() fails to crash the program. However, if you comment out line #2 *or* uncomment line #22, the program works as expected - that is, the assert catches the mistake.

You can compile with -debug to take a look at the returned data and this shows that it is being set correctly, its just that assert() doesn't seem to see the most recent value.

I chopped down this example from a 3,600 line program so its taken a long while to isolate the offending conditions.

-- 
Derek
Melbourne, Australia
31/05/2005 11:09:36 AM

Sorry about that. It'll get fixed next update.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Derek Parnell schrieb am Tue, 31 May 2005 11:19:42 +1000:
> Given the program below ...
>
> [test.d]
><code>
> debug(1) import std.stdio;
> const int anything = -1000; // Line #2
> dchar[] SomeFunc( dchar[] pText, out int pStopPosn)
> {
>     if (pText.length == 0)
>         pStopPosn = 0;
>     else
>         pStopPosn = -1;
>     debug(1) writefln("DEBUG: using '%s' we get %d", pText, pStopPosn);
>     return pText.dup;
> }
>
> int main(char[][] pArgs)
> {
>     int sp;
>
>     SomeFunc("123", sp);
>     debug(1) writefln("DEBUG: got %d", sp);
>     assert(sp == -1);
>
>     SomeFunc("", sp);
> //    if (sp != 0){} // Line #22
>     debug(1) writefln("DEBUG: got %d", sp);
>     assert(sp == -1);
>     return 0;
> }
></code>
>
> and compiled with "dmd test.d" version 0.125, the last assert() fails to crash the program. However, if you comment out line #2 *or* uncomment line #22, the program works as expected - that is, the assert catches the mistake.
>
> You can compile with -debug to take a look at the returned data and this shows that it is being set correctly, its just that assert() doesn't seem to see the most recent value.
>
> I chopped down this example from a 3,600 line program so its taken a long while to isolate the offending conditions.
>

Added to DStress as http://dstress.kuehne.cn/norun/a/assert_10_A.d http://dstress.kuehne.cn/norun/a/assert_10_B.d http://dstress.kuehne.cn/norun/a/assert_10_C.d http://dstress.kuehne.cn/norun/a/assert_10_D.d


Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFCnixn3w+/yD4P9tIRAocIAJ9IdbhBdGvodwr0ZzcCsKkKjGKnEACeOamV
0JBF7QyrAOSMGjs77jmiJqY=
=IuJ8
-----END PGP SIGNATURE-----