[llvm-commits] [llvm-gcc-4.2] r68848 - /llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp

[Devang Patel]

On Apr 14, 2009, at 10:11 AM, Dan Gohman wrote:

> On Apr 13, 2009, at 1:05 PM, Devang Patel wrote:
>
>>
>> On Apr 13, 2009, at 12:59 PM, Dan Gohman wrote:
>>
>>>
>>> On Apr 10, 2009, at 5:17 PM, Devang Patel wrote:
>>>
>>>> Author: dpatel
>>>> Date: Fri Apr 10 19:17:54 2009
>>>> New Revision: 68848
>>>>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=68848&view=rev
>>>> Log:
>>>> If linkage name is not available then use function display name as
>>>> linkage name.
>>>
>>> Hi Devang,
>>>
>>> What does this fix? Why is a bogus linkage name better than no
>>> linkage name?
>>
>>
>> In C, linkage name is same as name.
>
> And in pretty much everything else, it can be different.

And if it is different then gdb wants to know it. If linkage name is  
different then the FE should have found the linkage name.

>> This info is used during FastISel to decide whether we are at the
>> beginning of inlined subroutine or not.
>>
>>      std::string SPName;
>>      Subprogram.getLinkageName(SPName);
>>      if (!SPName.empty()
>>          && strcmp(SPName.c_str(), MF.getFunction()-
>>> getNameStart())) {
>
> Could this use
>
>    Subprogram.getGV() == MF.getFunction()
>
> instead?

It won't work. The Subprogram does not have actual Fn. It only carries  
information for function named "blah".

-
Devang

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20090414/a1bc13ec/attachment.html>