[cfe-dev] WIP initialization rewrite

[Steve Naroff]

Thanks for the reply.

I've started making a series of "small" changes (some of them  
influenced by your patch).

Given the complexity of initializers, I think it's best to keep  
everything working as we go.

snaroff

On Jan 21, 2008, at 7:25 PM, Eli Friedman wrote:

> Hmm, I can't check right now, but I did run the tests, and I'm pretty
> sure there weren't that many failures... I'll have to try against a
> clean tree to know for sure.  Off the top of my head, though, I'm
> pretty sure this patch depends on my string patch because it assumes
> strings are arrays.
>
> -Eli
>
> On Jan 21, 2008 10:28 AM, Steve Naroff <snaroff at apple.com> wrote:
>>
>> On Jan 19, 2008, at 8:56 AM, Eli Friedman wrote:
>>
>>
>>> I figured I should get this out into the open, since my holding onto
>>> it isn't very useful.  It's essentially a patch replacing the  
>>> whole of
>>> initialization-list processing sema with a version written from
>>> scratch.
>>>
>>> I'm not really asking for review of this, because it has quite a few
>>> weaknesses.
>>>
>>> 1. It rewrites the whole thing from scratch, which is probably not  
>>> the
>>> best idea, even if it is only a few hundred lines.
>>> 2. I'm not really satisfied with the way this patch is written; it  
>>> has
>>> a lot of code duplication. Some of the logic is a bit tricky, and  
>>> both
>>> the duplicate loops both within the method and the inablility to  
>>> reuse
>>> the logic for codegen/analysis are serious issues.
>>> 3. It's not complete: I haven't gone over any of the logic required
>>> for vectors.
>>>
>>> That said, the code does work, and I believe it implements C99 rules
>>> correctly, so it might be useful source of ideas for Steve or  
>>> whoever
>>> else touches this code.
>>
>> Eli,
>>
>> When I integrate your patch, I get a significant number of failures.
>>
>> Since you said the code works, I was surprised by the scope of the
>> breakage...
>>
>> Did you ever run these tests?
>>
>> Curious,
>>
>> snaroff
>>
>> [snaroff:tools/clang/test] snarofflocal% grep FAILED LOG
>> ******************** TEST 'CodeGen/global-with-initialiser.c' FAILED!
>> ********************
>> ******************** TEST 'CodeGen/globalinit.c' FAILED!
>> ********************
>> ******************** TEST 'CodeGen/ocu-vector.c' FAILED!
>> ********************
>> ******************** TEST 'CodeGen/shared-string-literals.c' FAILED!
>> ********************
>> ******************** TEST 'CodeGen/staticinit.c' FAILED!
>> ********************
>> ******************** TEST 'CodeGen/string-literal.c' FAILED!
>> ********************
>> ******************** TEST 'Parser/compound_literal.c' FAILED!
>> ********************
>> ******************** TEST 'Parser/ocu_vector_components.c' FAILED!
>> ********************
>> ******************** TEST 'Sema/array-constraint.c' FAILED!
>> ********************
>> ******************** TEST 'Sema/array-init.c' FAILED!
>> ********************
>> ******************** TEST 'Sema/carbon.c' FAILED!  
>> ********************
>> ******************** TEST 'Sema/cocoa.m' FAILED! ********************
>>
>>>
>>>
>>> -Eli
>>> <rewritewipdiff.txt>
>>
>>