[LLVMdev] Value of structure passed byval to a recurse function not initialized when accessed through GDB

[Karthik Bhat]

Hi,
There are few clang test cases which address this issue but are not
fixed. E.g. clang/test/CodeGen/x86_32-arguments-linux.c ,
tools/clang/test/CodeGen/x86_32-arguments-darwin.c etc have mentioned
them with a FIXME: note.

Will see if i can come up with some other examples and data.

Thanks

On Thu, Dec 6, 2012 at 10:55 PM, David Blaikie <dblaikie at gmail.com> wrote:
> On Thu, Dec 6, 2012 at 12:33 AM, Karthik Bhat <karthikthecool at gmail.com> wrote:
>> Hi David,
>>
>> I think it might not be exactly PR13303 which might be causing the
>> corruption of struct when accessed through GDB.
>> This seems to be an ABI problem in clang.
>> The problem seems to be that when we have pass by value of struct
>> (having indirect arguments) stack is not aligned properly.
>>
>> I tried realigning the stack for indirect arguments in(TargetInfo.cpp) -
>>
>> ABIArgInfo X86_32ABIInfo::getIndirectResult(QualType Ty, bool ByVal)
>>
>> .....
>>  if (StackAlign == 0)
>>         return ABIArgInfo::getIndirect(4, /*ByVal=*/true,
>>                                    /*Realign=*/true);      // Do a
>> realign of stack.
>>
>> ...
>>
>>
>> This seems to have fixed the issue. Also in case we have a large
>> structure - e.g. -
>>
>> typedef struct s
>> {
>>   long s;
>>   long i;
>>   long l;
>>   long s1;
>>   long i1;
>>   long l1;
>> } SVAL;
>>
>> in the above mentioned code the same issue(corruption of member
>> variables when accessed through GDB) was observed which has got fixed
>> after this change.
>>
>> Need input if this change is correct.
>
> I haven't looked at this carefully yet (& I'm not the authority on ABI
> issues - not sure who pays most attention to this in the backend, John
> McCall deals with it mostly in Clang proper but I've CC'd him here in
> case things rings any bells for him) but a simple way you could
> provide strong motivation for this change is if you can demonstrate
> that this is also a correctness issue: If clang/llvm are really
> incorrectly implementing the ABI this should cause interoperability
> issues if clang is used to compile a caller and gcc a callee to the
> same function (or the other way around).
>
> If you can come up with a simple example that demonstrates that it
> should be a fairly unquestionable change. (test cases along with the
> change all in a patch file help too)
>
> Thanks,
> - David
>
>>
>> Thanks
>>
>> On Wed, Dec 5, 2012 at 1:45 AM, David Blaikie <dblaikie at gmail.com> wrote:
>>> This seems to be another case of PR13303 - since GDB can't figure out
>>> where to break for this function based on the debug info (you'll
>>> notice when you "break recurse" that it's not breaking on a line or
>>> source file, just an address) it's breaking at the very start, before
>>> the prologue
>>>
>>> I'm about to commit a fix to this.
>>>
>>> On Tue, Dec 4, 2012 at 5:34 AM, Karthik Bhat <karthikthecool at gmail.com> wrote:
>>>> Hi All,
>>>>
>>>> I was debugging a clang binary when i found this problem. The
>>>> following code is complied with clang.
>>>>
>>>> typedef struct s
>>>> {
>>>>   short s;
>>>> } SVAL;
>>>>
>>>>
>>>> void recurse (SVAL a, int depth)
>>>> {
>>>>   a.s = --depth;
>>>>   if (depth == 0)
>>>>     return;
>>>>   else
>>>>    recurse(a,depth);
>>>> }
>>>>
>>>> int main ()
>>>> {
>>>>   SVAL s; s.s = 5;
>>>>   recurse (s, 5);
>>>>   return 0;
>>>> }
>>>>
>>>> When i try to access value of a.s in function recurse through gdb(i.e
>>>> gdb > p a.s) it gives me an uninitialized value.
>>>> The problem occurs only when we have a function call within function
>>>> to which we have passed a structure.
>>>>
>>>> Could someone guide me were can i look to fix this issue.
>>>>
>>>> I have started with LowerFormalArguments in X86ISelLowering.cpp file.
>>>>
>>>> Thanks
>>>> Karthik
>>>> _______________________________________________
>>>> LLVM Developers mailing list
>>>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev