[cfe-commits] r124507 - in /cfe/trunk: lib/CodeGen/CGCXX.cpp lib/CodeGen/CGExprCXX.cpp lib/CodeGen/CGVTables.cpp test/CodeGenCXX/apple-kext-indirect-call-2.C
Douglas Gregor
dgregor at apple.com
Mon Jan 31 09:45:40 PST 2011
On Jan 31, 2011, at 9:17 AM, jahanian wrote:
>
> On Jan 31, 2011, at 7:26 AM, Douglas Gregor wrote:
>
>>
>> On Jan 28, 2011, at 3:42 PM, Fariborz Jahanian wrote:
>>
>>> Author: fjahanian
>>> Date: Fri Jan 28 17:42:29 2011
>>> New Revision: 124507
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=124507&view=rev
>>> Log:
>>> More work to support -fapple-kext regarding
>>> indirect vf calls and addition of extra entry
>>> at bottom of vtbls.
>>>
>>> Added:
>>> cfe/trunk/test/CodeGenCXX/apple-kext-indirect-call-2.C
>>> Modified:
>>> cfe/trunk/lib/CodeGen/CGCXX.cpp
>>> cfe/trunk/lib/CodeGen/CGExprCXX.cpp
>>> cfe/trunk/lib/CodeGen/CGVTables.cpp
>>>
>>> Modified: cfe/trunk/lib/CodeGen/CGCXX.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGCXX.cpp?rev=124507&r1=124506&r2=124507&view=diff
>>> ==============================================================================
>>> --- cfe/trunk/lib/CodeGen/CGCXX.cpp (original)
>>> +++ cfe/trunk/lib/CodeGen/CGCXX.cpp Fri Jan 28 17:42:29 2011
>>> @@ -329,7 +329,9 @@
>>> assert(VTable && "BuildVirtualCall = kext vtbl pointer is null");
>>> MD = MD->getCanonicalDecl();
>>> uint64_t VTableIndex = CGM.getVTables().getMethodVTableIndex(MD);
>>> - VTableIndex += 2;
>>> + uint64_t AddressPoint =
>>> + CGM.getVTables().getAddressPoint(BaseSubobject(RD, 0), RD);
>>> + VTableIndex += AddressPoint;
>>> llvm::Value *VFuncPtr =
>>> CGF.Builder.CreateConstInBoundsGEP1_64(VTable, VTableIndex, "vfnkxt");
>>> return CGF.Builder.CreateLoad(VFuncPtr);
>>
>> Ah, I think I understand this now. We'll want to document what we're doing somewhere.
>
> Yes. We have have the documentation of what -fapple-kext does after al pieces are known and are in place.
Okay!
>>
>>> Modified: cfe/trunk/lib/CodeGen/CGVTables.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGVTables.cpp?rev=124507&r1=124506&r2=124507&view=diff
>>> ==============================================================================
>>> --- cfe/trunk/lib/CodeGen/CGVTables.cpp (original)
>>> +++ cfe/trunk/lib/CodeGen/CGVTables.cpp Fri Jan 28 17:42:29 2011
>>> @@ -2758,7 +2758,14 @@
>>>
>>> // Add the VTable layout.
>>> uint64_t NumVTableComponents = Builder.getNumVTableComponents();
>>> + // -fapple-kext adds an extra entry at end of vtbl.
>>> + bool IsAppleKext = CGM.getContext().getLangOptions().AppleKext;
>>> + if (IsAppleKext)
>>> + NumVTableComponents += 1;
>>> +
>>> uint64_t *LayoutData = new uint64_t[NumVTableComponents + 1];
>>> + if (IsAppleKext)
>>> + LayoutData[NumVTableComponents] = 0;
>>> Entry.setPointer(LayoutData);
>>
>> NumVTableComponents-1?
>
> We are adding a null entry after all other entries. So, last entry is nullified.
Oh, of course, sorry. I blame Monday!
- Doug
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110131/ed5bcc44/attachment.html>
More information about the cfe-commits
mailing list