This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH]: Fix problematic -Wcast-qual cases using new CONST_CAST macro
- From: "Kaveh R. GHAZI" <ghazi at caip dot rutgers dot edu>
- To: Richard Guenther <richard dot guenther at gmail dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 3 Aug 2007 10:19:47 -0400 (EDT)
- Subject: Re: [PATCH]: Fix problematic -Wcast-qual cases using new CONST_CAST macro
- References: <Pine.GSO.4.58.0707270359140.28643@caipclassic.rutgers.edu> <84fc9c000707270219i4702de61y70899cd7e3349d72@mail.gmail.com> <Pine.GSO.4.58.0707271249520.7916@caipclassic.rutgers.edu> <84fc9c000707271138u6171f592x3f69800f12106a0a@mail.gmail.com> <Pine.GSO.4.58.0707271509350.16274@caipclassic.rutgers.edu>
On Fri, 27 Jul 2007, Kaveh R. GHAZI wrote:
> On Fri, 27 Jul 2007, Richard Guenther wrote:
>
> > On 7/27/07, Kaveh R. GHAZI <ghazi@caip.rutgers.edu> wrote:
> > > Sure. Besides my main point which is silencing -Wcast-qual warnings, the
> > > CONST_CAST macro is grep-able. I.e. you can search for the few places
> > > where it's used and audit if they are correctly applied. You cannot do
> > > that with regular C-style casts. Another reason is that with C-style
> > > casts there isn't any indication of why it's there. The CONST_CAST macro
> > > comments itself to someone reading the code. (These are some of the
> > > rationales for the creation of C++'s cast operators.) The macro isn't
> > > type-safe like the C++ operator is, but it's a poor man's C option. :-)
> >
> > Ha! Maybe a reason to switch to C++ ;) How about
> >
> > #ifdef __cplusplus
> > #define CONST_CAST(x) .. template magic follows ..
> > #else
> > ...
> > #endif
> >
> > ? ;)
> > Richard.
>
> I don't think that will work in a generic way. For static_cast, you need
> to supply the stripped type like so:
>
> static_cast<STRIPPED-TYPE>(EXPR)
>
> Since we don't yet have a variant of __typeof to get the qualifer stripped
> type for us, we'd need to pass in the stripped type to the macro like so:
>
> #ifdef __cplusplus
> #define CONST_CAST(X,TYPE) static_cast<TYPE>(X)
>
> Note this doesn't increase type safety for most of the cases we'll use it
> in because we're just passing the pointer into a situation that takes a
> void* like free(foo). I'm not sure this change is worth it so I'd rather
> leave this refinement to the folks who care about compiling gcc with a C++
> compiler if they care to take this step.
>
> Also, I don't believe the main sources are compilable with a C++ compiler
> at this point anyway. Only the libiberty/ and gcc/cp/ directories use
> -Wc++-compat AFAICT.
>
> So may I install the latest version of my patch?
Richard, do you have any remaining objections to this patch?
http://gcc.gnu.org/ml/gcc-patches/2007-07/msg01993.html
If there is some C++ magic you want, can you please suggest something?
I.e. do you want me to add a TYPE parameter to CONST_CAST? Perhaps
CONST_CAST_WITH_TYPE? Otherwise I'd like to install it.
Thanks,
--Kaveh
--
Kaveh R. Ghazi ghazi@caip.rutgers.edu