[cfe-dev] Modifying an AST and Sema's dependency on an AST consumer

[Ted Kremenek]

On Dec 4, 2008, at 8:44 AM, Daniel Dunbar <daniel at zuster.org> wrote:

> However, it is perfectly reasonable for a client to want to use ASTs
> and delete the Preprocessor.

Absolutely.

> Does this mean we should have another
> context object?

That's a possibilty, but what would such a context object represent?   
My main concern is that at some point a proliferation of such objects  
becomes more cumbersome than the benefit they provide.

> On Thu, Dec 4, 2008 at 8:27 AM, Ted Kremenek <kremenek at apple.com>  
> wrote:
>> On Dec 4, 2008, at 7:26 AM, steve naroff <snaroff at apple.com> wrote:
>>
>>>
>>> On Dec 4, 2008, at 9:54 AM, Lukasz Janyst wrote:
>>>
>>>> On Wed, Dec 3, 2008 at 3:26 PM, Lukasz Janyst <ljanyst at cern.ch>
>>>> wrote:
>>>>> tables in it. Yet still I seem to experience some strange memory
>>>>> issues when the modified TU gets processed by the CodeGen.
>>>>
>>>> I sorted it out. The problem was the object ownership. It seemed
>>>> natural to me that the identifier table should be owned by the
>>>> ASTContext, but it is owned by the Preprocessor object which I
>>>> deleted
>>>> after I was done with the parsing.
>>>>
>>>
>>> The IdentifierTable should be owned by ASTContext (as you expected).
>>>
>>> This is a bug (that I thought we fixed a long time ago...apparently
>>> not).
>>>
>>> snaroff
>>
>> I don't think this is a bug.  See my other email.
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>