[cfe-commits] r164874 - in /cfe/trunk: include/clang/Lex/PPCallbacks.h include/clang/Lex/PreprocessingRecord.h lib/Frontend/DependencyFile.cpp lib/Frontend/DependencyGraph.cpp lib/Lex/PPDirectives.cpp lib/Lex/PreprocessingRecord.cpp lib/Lex/Prepr

[Chandler Carruth]

On Mon, Oct 1, 2012 at 10:51 AM, Argyrios Kyrtzidis <kyrtzidis at apple.com>wrote:

> Hi Kim,
>
> On Sep 30, 2012, at 12:29 AM, Kim Gräsman <kim.grasman at gmail.com> wrote:
>
> > On Sat, Sep 29, 2012 at 11:50 PM, Kim Gräsman <kim.grasman at gmail.com>
> wrote:
> >>
> >> I would expect either '<stdio.h>' or
> >> 'HEADER_NAME', but I think getting the expanded range ('<stdio.h>')
> >> would be the most natural.
> >
> > Having looked into our callback implementation, we already handle
> > macro expansion, and I think passing down the macro name range
> > unexpanded gives more fidelity. So I change my mind -- I think getting
> > the range as-written would be better ('HEADER_NAME'). Of course,
> > assuming that FilenameRange.getBegin().isMacroID() is still true so
> > callback sinks can expand it.
>
> Your suggestion is a bit contradictory, if we get the range 'HEADER_NAME',
> then FilenameRange.getBegin().isMacroID() will be false (since it's
> pointing at the beginning of the macro expansion).
> The most fidelity option would be to give a range for  '<stdio.h>'
> (MacroIDs for begin and end), from which a callback sink can get at the
> 'HEADER_NAME' file-level range.
> Is this your recommendation ?
>

FWIW, I think it is, and it makes sense. ;] The source location structure
is... confusing though.

Richard, thoughts?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20121001/0c0263ca/attachment.html>