Atomic increment only works for bytes

I don't understand your problem, inc [eax] is used only for 1 byte increments (as far as I can see), in which case it is perfectly ok to do incb [eax].

If I missed something please reopen and give more context (like line number).

tango.core.Atomic.atomicIncrement should increment integer value up to int.max and then drop to 0.

It only increments the least significant byte as of now.

Here is a test case:

int i = 255; atomicIncrement(i); assert(i == 256); // fails

Err, I meant "...and then drop to int.min", not to 0.

oh, you're saying the assembler should be something like:

inc dword ptr [eax];

?

Exactly. But it should be int ptr, not dword ptr. (Weird, I know).

well it *is*

inc int ptr [eax];

as far as I can see, and indeed the test case works for me on gdc/mac and dmd 1.039/ubuntu

If as Don says int ptr is correct (and Don is an asm expert, so I think it is so), I would say that looks like a compiler bug, can you give more info about your platform/compiler version korDen?

add also the tango version to that...

I just checked on codepad.org and indeed the example fails on it. They use Digital Mars D Compiler v1.026 and tango 0.99 I strongly suspect that with the latest versions it is fixed...

Apparently, I was using an outdated version of Tango and the bug is already fixed per ticket #1221 in changeset [3929].

jhouse: can you confirm that your issue is fixed?

It appears that's the right fix. It's unpatched in the D2.0 branch of Tango and exists in the 0.99.7 release. I've updated the version on the ticket.

I wish I was automatically cc'd on changes to tickets I submit. I forgot to add myself to the cc list when I submitted the bug report.