[PATCH] [ms-cxxabi] Correctly compute the size of member pointers

[Reid Kleckner]

On Thu, Mar 21, 2013 at 9:17 PM, João Matos <ripzonetriton at gmail.com> wrote:

> On Thu, Mar 21, 2013 at 11:22 PM, Will Wilson <will at indefiant.com> wrote:
>
>> Oddly enough I've been working on this exact area myself over the last
>> few days. I could be overlooking something but it appears to be missing
>> support for the X64 ABI (ie. amd64-pc-win32) which would be a shame not to
>> support. I've previously spent some time analyzing it and came up with the
>> following member function pointer sizes:
>>
>
> I agree, X64 ABI support would be nice.
>

It would be nice, but I'd rather have i386 first.  :)


> The inheritance keywords should be definitive for the rest of the TU. MSVC
>> will issue an error should the class definition not match the assigned
>> inheritance keyword - this is a useful sanity check and probably worth
>> implementing. The error in question is C2286 and is outlined here:
>> http://msdn.microsoft.com/en-gb/library/9z8tbhb2.aspx
>>
>
> I already tried to add a warning for this, see r163078, but that got
> reverted at the time. We should definitely add a warning for this.
>

Yeah, I didn't add that into this patch to focus on just the ABI stuff.

In addition, MSVC has a nasty corner case. If a member function pointer
>> is instantiated based on an (as yet) undefined class the class is assigned
>> the unknown inheritance for the rest of the translation unit - even if the
>> class is defined later in the TU. I've been wondering how best to model
>> this persistant state and the only approach I could come up with was to
>> assign a (currently unimplemented) UnknownInheritanceAttr during sema - but
>> it's not pretty.
>>
>
> Interesting, that sounds indeed a bit tricky to implement.
>
>
>> Lastly, I haven't had a chance to revisit it, but MSVC seems to do weird
>> things with member function pointer alignment. I know this isn't specific
>> to this patch but I thought I'd mention it. Hopefully I'll get the chance
>> to research it soon and get to the bottom of the implementation.
>>
>
>  File a bug about this when you know more details so we do not forget it.
>
> --
> João Matos
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130321/02265722/attachment.html>