Just did a checkin. My sources are not different.

Don Clugston wrote:
> It fails for me, too. Your version must be different to the version in svn. See also the ICE regression bug 3726 which popped up in the most recent commit.
>
>
> 2010/1/21 Walter Bright <walter at digitalmars.com>:
> 
>> I tried it again, it still works. All I can suggest is you are working from a different set of dmd sources than I am.
>>
>> Rainer Schuetze wrote:
>> 
>>> Hi,
>>>
>>> I just tried again, rebuilt dmd after clean, but the error persists
>>> (though it's different from the one reported in the bugzilla entry #3674). I
>>> have added a detailed description to the entry.
>>>
>>> I think this has crawled into dmd with changes to the Type::covariant()
>>> function, but I have no clue why it didn't show up earlier (and why you
>>> don't see it).
>>>
>>> Rainer
>>>
>>> Walter Bright wrote:
>>> 
>>>> I just tried it, and it works.
>>>>
>>>> Rainer Schuetze wrote:
>>>> 
>>>>> Hi,
>>>>>
>>>>> I just wanted to drop a note that there seems to be regression with the latest changes committed to subversion. It seems to be related to #3674 (the test-case fails), but I have no clue yet what's been causing this. I will try to figure out what's going on tonight.
>>>>>
>>>>> Rainer
>>>>>
>>>>> Rainer Schuetze wrote:
>>>>> 
>>>>>> Hi,
>>>>>>
>>>>>> Walter Bright wrote:
>>>>>> 
>>>>>>> Rainer Schuetze wrote:
>>>>>>> 
>>>>>>>> So, if you want to make the qtd-people happy, issues #3600 and #3647
>>>>>>>> should be addressed.
>>>>>>>> 
>>>>>>> They're done now.
>>>>>>>
>>>>>>> 
>>>>>> Cool, thanks.
>>>>>>
>>>>>> 
>>>>>>>> And another thing: when there are forward references that are not
>>>>>>>> resolved for a long time, Module::runDeferredSemantic() can waste a lot of
>>>>>>>> memory and cpu-cycles. It should not retry between every new symbol in a
>>>>>>>> class or struct, but only when some sensible progress is made (e.g. if
>>>>>>>> Module::dprogress > 0). runDeferredSemantic() must then be forced to run
>>>>>>>> once before running semantic2() on the modules.
>>>>>>>> It's not easy to create a small test case, but qtd fails to compile
>>>>>>>> with "out of memory" without this change.
>>>>>>>>
>>>>>>>> 
>>>>>>> Could you put your patch into bugzilla, please?
>>>>>>> 
>>>>>> It can now be found here:
>>>>>>
>>>>>> http://d.puremagic.com/issues/show_bug.cgi?id=3719
>>>>>>
>>>>>> Rainer
>>>>>>
>>>>>> _______________________________________________
>>>>>> dmd-beta mailing list
>>>>>> dmd-beta at puremagic.com
>>>>>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>>>>>
>>>>>> 
>>>>> _______________________________________________
>>>>> dmd-beta mailing list
>>>>> dmd-beta at puremagic.com
>>>>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>>>>
>>>>>
>>>>> 
>>>> _______________________________________________
>>>> dmd-beta mailing list
>>>> dmd-beta at puremagic.com
>>>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>>>
>>>> 
>>> _______________________________________________
>>> dmd-beta mailing list
>>> dmd-beta at puremagic.com
>>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>>
>>>
>>> 
>> _______________________________________________
>> dmd-beta mailing list
>> dmd-beta at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>
>> 
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>
>
> 

An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/dmd-beta/attachments/20100121/fa52faa8/attachment.htm>

Checked in changes to fix second problem with 3674.

Rainer Schuetze wrote:
> Hi,
>
> sorry for causing confusion, but it still won't compile here. Getting a fresh copy from the repository showed no difference in the source code. I've also tried going back to revision 324 with the original patch, still an error. qtd compiled a week ago, so I'm completely confused.
>
> Do you see a flaw in the analysis in bugzilla #3674? Maybe discussion should be done there...
>
> Rainer
>
> Walter Bright wrote:
>> I tried it again, it still works. All I can suggest is you are working from a different set of dmd sources than I am.
>>
>> Rainer Schuetze wrote:
>>> Hi,
>>>
>>> I just tried again, rebuilt dmd after clean, but the error persists (though it's different from the one reported in the bugzilla entry #3674). I have added a detailed description to the entry.
>>>
>>> I think this has crawled into dmd with changes to the Type::covariant() function, but I have no clue why it didn't show up earlier (and why you don't see it).
>>>
>>> Rainer
>>>
>>> Walter Bright wrote:
>>>> I just tried it, and it works.
>>>>
>>>> Rainer Schuetze wrote:
>>>>> Hi,
>>>>>
>>>>> I just wanted to drop a note that there seems to be regression with the latest changes committed to subversion. It seems to be related to #3674 (the test-case fails), but I have no clue yet what's been causing this. I will try to figure out what's going on tonight.
>>>>>
>>>>> Rainer
>
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>
>

Don Clugston wrote:
> See also the ICE regression bug 3726 which popped up in the most recent commit.
>
> 

Done.

To be honest, I thought the test cases were almost the same, just one class removed. I noticed the difference "class" <-> "interface" just a minute ago - this must have crept in earlier when trying the patch against different variations.

sorry for the trouble and thanks for your patience,
Rainer

Walter Bright wrote:
> Ok, I see the misunderstanding now. The original test case in bugzilla compiles and works fine. You added a /different/ test case which fails in a different way.
>
> Rainer Schuetze wrote:
>> Hi,
>>
>> sorry for causing confusion, but it still won't compile here. Getting a fresh copy from the repository showed no difference in the source code. I've also tried going back to revision 324 with the original patch, still an error. qtd compiled a week ago, so I'm completely confused.
>>
>> Do you see a flaw in the analysis in bugzilla #3674? Maybe discussion should be done there...
>>
>> Rainer
>>
>> Walter Bright wrote:
>>> I tried it again, it still works. All I can suggest is you are working from a different set of dmd sources than I am.
>>>
>>> Rainer Schuetze wrote:
>>>> Hi,
>>>>
>>>> I just tried again, rebuilt dmd after clean, but the error persists (though it's different from the one reported in the bugzilla entry #3674). I have added a detailed description to the entry.
>>>>
>>>> I think this has crawled into dmd with changes to the Type::covariant() function, but I have no clue why it didn't show up earlier (and why you don't see it).
>>>>
>>>> Rainer
>>>>
>>>> Walter Bright wrote:
>>>>> I just tried it, and it works.
>>>>>
>>>>> Rainer Schuetze wrote:
>>>>>> Hi,
>>>>>>
>>>>>> I just wanted to drop a note that there seems to be regression with the latest changes committed to subversion. It seems to be related to #3674 (the test-case fails), but I have no clue yet what's been causing this. I will try to figure out what's going on tonight.
>>>>>>
>>>>>> Rainer
>>
>> _______________________________________________
>> dmd-beta mailing list
>> dmd-beta at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>
>>
> ------------------------------------------------------------------------
>
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta

It's no trouble, I'm just glad to get it fixed. It's in the test suite now.

Rainer Schuetze wrote:
> To be honest, I thought the test cases were almost the same, just one class removed. I noticed the difference "class" <-> "interface" just a minute ago - this must have crept in earlier when trying the patch against different variations.
>
> sorry for the trouble and thanks for your patience,
> Rainer
>

Walter Bright, el 22 de enero a las 01:10 me escribiste:
> It's no trouble, I'm just glad to get it fixed. It's in the test suite now.

It just occurred to me that, since your test suite have a lot of collected code from multiple unknown source you can't distribute it freely. What about having the test suite splat in 2, one with the non-distributable test cases and one with the distributable ones. This way, new test cases can be added to the test free test suite and over the time, the free test suite could grow in size to the point to be really useful for other implementations or for people wanting to develop DMD without breaking stuff :)

You can even add some clause to the bugzilla stating that any code published there should be Public Domain (or Boost licenced or whatever fits the purpose) so you can be sure you can grab any test case in bugzilla from now on.

What do you think?

-- 
Leandro Lucarella (AKA luca)                     http://llucax.com.ar/
----------------------------------------------------------------------
GPG Key: 5F5A8D05 (F8CD F9A7 BF00 5431 4145  104C 949E BFB6 5F5A 8D05)
----------------------------------------------------------------------

Leandro Lucarella wrote:
> Walter Bright, el 22 de enero a las 01:10 me escribiste:
> 
>> It's no trouble, I'm just glad to get it fixed. It's in the test suite now.
>> 
>
> It just occurred to me that, since your test suite have a lot of collected code from multiple unknown source you can't distribute it freely. What about having the test suite splat in 2, one with the non-distributable test cases and one with the distributable ones. This way, new test cases can be added to the test free test suite and over the time, the free test suite could grow in size to the point to be really useful for other implementations or for people wanting to develop DMD without breaking stuff :)
>
> You can even add some clause to the bugzilla stating that any code published there should be Public Domain (or Boost licenced or whatever fits the purpose) so you can be sure you can grab any test case in bugzilla from now on.
>
> What do you think?
> 

It's a good idea.