r204846 - InstrProf: Use unique_ptr

[Duncan P. N. Exon Smith]

On Mar 26, 2014, at 12:36 PM, David Blaikie <dblaikie at gmail.com> wrote:

> On Wed, Mar 26, 2014 at 12:26 PM, Duncan P. N. Exon Smith
> <dexonsmith at apple.com> wrote:
>> Author: dexonsmith
>> Date: Wed Mar 26 14:26:05 2014
>> New Revision: 204846
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=204846&view=rev
>> Log:
>> InstrProf: Use unique_ptr
>> 
>> Modified:
>>    cfe/trunk/lib/CodeGen/CodeGenPGO.cpp
>>    cfe/trunk/lib/CodeGen/CodeGenPGO.h
>> 
>> Modified: cfe/trunk/lib/CodeGen/CodeGenPGO.cpp
>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CodeGenPGO.cpp?rev=204846&r1=204845&r2=204846&view=diff
>> ==============================================================================
>> --- cfe/trunk/lib/CodeGen/CodeGenPGO.cpp (original)
>> +++ cfe/trunk/lib/CodeGen/CodeGenPGO.cpp Wed Mar 26 14:26:05 2014
>> @@ -162,7 +162,7 @@ void CodeGenPGO::setFuncName(llvm::Funct
>>     RawFuncName = RawFuncName.substr(1);
>> 
>>   if (!Fn->hasLocalLinkage()) {
>> -    PrefixedFuncName = new std::string(RawFuncName);
>> +    PrefixedFuncName.reset(new std::string(RawFuncName));
> 
> Um - why do you have a new'd std::string? Any reason you can't just
> have a straight value-std::string and use the empty string to
> represent "not present" (if that's what the unique_ptr is for - to
> allow null to represent the "not a string" case?)
> 
> (or, if you need "not a string" and "empty string" to be separate
> concepts - Optional<std::string> might be appropriate)

I’ve been wondering that myself since I found it :).  There’s a series of
containers in CodeGenPGO that use new that probably needn’t (the same
series that I just converted to unique_ptr).  On my list of things to clean
up when I get a chance.