Re: [C++ Patch/RFC] PR 54194

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi again,

On 10/08/2012 11:44 PM, Jason Merrill wrote:
> On 10/08/2012 05:31 PM, Paolo Carlini wrote:
> > So, there is a serious difficulty, I'm afraid: for the example at issue,
> > EXPR_LOCATION (arg_left) is 0 not any meaningful value. And of course
> > EXPR_LOC_OR_HERE would not be better in this case, would give
> > input_location. So, what do you think? Shall we just go with the loc for
> > now, or you think there is something relatively safe we can try?
> Looks like we still need to SET_EXPR_LOCATION in cp_build_binary_op.  
> It would also be good to have a macro/function like EXPR_LOC_OR_HERE 
> that lets you provide the location to use if the expression doesn't 
> have one.
Thus I'm finishing testing the below (C++ is Ok)

It works pretty well, IMHO: for actual expressions we can have the 
column number of the operand (which is now fixed up in 
cp_parser_binary_expression), otherwise, for eg constants, we fall back 
to the column number of the operator. As a bonus, in templates like 
Wparentheses-26.C we even get the line number right.

Thanks!
Paolo.

////////////////////////////

Attachment: CL_54194
Description: Text document

Attachment: patch_54194
Description: Text document