[cfe-commits] r73821 - in /cfe/trunk: include/clang/AST/Decl.h include/clang/AST/DeclBase.h lib/AST/Decl.cpp

[Argyrios Kyrtzidis]

On Jun 22, 2009, at 9:37 AM, Douglas Gregor wrote:

>
>>
>> +void FunctionDecl::setBody(Stmt *B) {
>> +  Body = B;
>> +  if (B && EndRangeLoc < B->getLocEnd())
>> +    EndRangeLoc = B->getLocEnd();
>> +}
>
> Why do we need this < check? The body of the function definition will
> always have an end location later in the file than the last function
> parameter.

Yep, this is redundant, no matter what value EndRangeLoc had before,  
it should be updated here. I removed the check.

>
> I was about to point out that EndRangeLoc is unnecessary, because we
> could compute the full range in getSourceRange(). However, when we're
> dealing with a function definition deserialized from a PCH/AST file,
> we can only compute the full range once the function body has been de-
> serialized... so it's far better to have the (sometimes-redundant)
> EndRangeLoc. Please add a comment to that effect.

Done.

>
>> bool FunctionDecl::isMain() const {
>>  return getDeclContext()->getLookupContext()->isTranslationUnit() &&
>>    getIdentifier() && getIdentifier()->isStr("main");
>> @@ -481,6 +493,10 @@
>>    void *Mem = C.Allocate(sizeof(ParmVarDecl*)*NumParams);
>>    ParamInfo = new (Mem) ParmVarDecl*[NumParams];
>>    memcpy(ParamInfo, NewParamInfo, sizeof(ParmVarDecl*)*NumParams);
>> +
>> +    // Update source range.
>> +    if (EndRangeLoc < NewParamInfo[NumParams-1]->getLocEnd())
>> +      EndRangeLoc = NewParamInfo[NumParams-1]->getLocEnd();
>>  }
>> }
>
>
> I don't think we need to use the < here.

This is to allow the flexibility to set EndRangeLoc before setting the  
parameters.
i.e. at PCHDeclReader::VisitFunctionDecl, EndRangeLoc is set before  
the setParams() call. If the check was missing we would have to be  
careful about the exact order of the calls to "reconstruct" the  
FunctionDecl.


-Argiris

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20090622/bedac327/attachment.html>