Jacob Carlborg <doob@me.com> writes:

> On 2016-02-28 09:46, Dan Olson wrote:
>
>> Well, it isn't pretty but can pass the dmd testsuite for objc.  I'll organize it a litte better, get a pull going to show work in progress.
>
> That was quick :). Please let me know when you have a pull request available, I can help with the code review.
>
> Would be inserting to enable all tests in objc_objc_msgSend.d and see if they work in LDC.

I uncommented lines in main (I think that is what you mean) to this:

void main()
{
    test_stret();
    test_fp2ret();
    test_fpret();
    test_float32();
    test_double64();
}

And it works.  I was unsure if the fp2ret and C++/D complex types would work, but for this test they do.

Hi Kai !
I wonder if there plans to convert driver and gen parts into d too, to get rid of c++ code.
Thanks for a reply.

On Sunday, 28 February 2016 at 21:54:22 UTC, Temtaime wrote:
> I wonder if there plans to convert driver and gen parts into d too, to get rid of c++ code.

Some code has already been converted to D (out of necessity / ease of transitioning to DDMD). At least for code that does not interface with LLVM, it'd be nice to convert it to D (imho).

cheers,
  Johan

On 2016-02-28 20:33, Dan Olson wrote:

> I uncommented lines in main (I think that is what you mean) to this:
>
> void main()
> {
>      test_stret();
>      test_fp2ret();
>      test_fpret();
>      test_float32();
>      test_double64();
> }
>
> And it works.  I was unsure if the fp2ret and C++/D complex types would
> work, but for this test they do.

Awesome :)

-- 
/Jacob Carlborg

On Friday, 26 February 2016 at 21:12:43 UTC, Kai Nacke wrote:
> Again, we need to help to test this alpha release! Every feedback is welcomed!


It takes me a lot to understand what's wrong in a parser code.
I finally reduced code to this:

size_t inc(size_t* v)
{
   (*v)++;
   return 10;
}

size_t pos = 10;
	
immutable tst = pos + inc(&pos); // <-- this
writeln(tst);
writeln(pos);

The marked line was an unintended mistake inside original code.
Anyway it gives different result with dmd and ldc2 1.0. I neither know if is a undefined behaviour.

On Monday, 29 February 2016 at 11:51:31 UTC, Andrea Fontana wrote:
> It takes me a lot to understand what's wrong in a parser code.
> I finally reduced code to this:
>
> size_t inc(size_t* v)
> {
>    (*v)++;
>    return 10;
> }
>
> size_t pos = 10;
> 	
> immutable tst = pos + inc(&pos); // <-- this
> writeln(tst);
> writeln(pos);
>
> The marked line was an unintended mistake inside original code.
> Anyway it gives different result with dmd and ldc2 1.0. I neither know if is a undefined behaviour.

I hope LDC computes tst = 20. ;)
Iirc, there were some inconsistencies wrt. this. LDC and GDC do it one way (the proper one imo), dmd another way.

On Monday, 29 February 2016 at 13:36:59 UTC, kink wrote:
> I hope LDC computes tst = 20. ;)
> Iirc, there were some inconsistencies wrt. this. LDC and GDC do it one way (the proper one imo), dmd another way.

"Proper" depends on specs. If is an undefined behaviour i think it should be defined or a warning should be triggered

On Monday, 29 February 2016 at 11:51:31 UTC, Andrea Fontana wrote:
> On Friday, 26 February 2016 at 21:12:43 UTC, Kai Nacke wrote:
>> Again, we need to help to test this alpha release! Every feedback is welcomed!
>
>
> It takes me a lot to understand what's wrong in a parser code.

Another problem. Reduced code:
http://dpaste.dzfl.pl/33434c9ba374

This print "LEN:3" with dmd and "LEN:0" with ldc2 1.0.  (using online compiler ldc 0.12, it prints LEN:3)

On Monday, 29 February 2016 at 14:49:44 UTC, Andrea Fontana wrote:
> On Monday, 29 February 2016 at 11:51:31 UTC, Andrea Fontana wrote:
>> On Friday, 26 February 2016 at 21:12:43 UTC, Kai Nacke wrote:
>>> Again, we need to help to test this alpha release! Every feedback is welcomed!
>>
>>
>> It takes me a lot to understand what's wrong in a parser code.
>
> Another problem. Reduced code:
> http://dpaste.dzfl.pl/33434c9ba374
>
> This print "LEN:3" with dmd and "LEN:0" with ldc2 1.0.  (using online compiler ldc 0.12, it prints LEN:3)

(please notice that if "auto" is used instead of "enum", len is computed correctly)

On Monday, 29 February 2016 at 14:47:58 UTC, Andrea Fontana wrote:
> On Monday, 29 February 2016 at 13:36:59 UTC, kink wrote:
>> I hope LDC computes tst = 20. ;)
>> Iirc, there were some inconsistencies wrt. this. LDC and GDC do it one way (the proper one imo), dmd another way.
>
> "Proper" depends on specs. If is an undefined behaviour i think it should be defined or a warning should be triggered

Spec:
https://dlang.org/spec/expression.html#order-of-evaluation

Strict left-to-right order evaluation, as the spec prescribes, would give tst=20.
Is there a DMD bug report for this? I didn't test with DMD, but if tst!=20, it is a pretty severe issue.