This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: V3 and Integrated CPP [was Re: V3 PATCH: Disable wide-character tests where necessary]


Jamie Lokier <egcs@tantalophile.demon.co.uk> writes:

> Neil Booth wrote:
> > Diagnostics produced by CPP will be produced almost exclusively during
> > the first pass; so will be at the right place.  The reread will have
> > incorrect column numbers, but cc1 has always had that problem.
> > 
> > Integrated cc1 will, in the future, be able to get the correct column
> > numbers for diagnostics (except in the --save-temps case).
> 
> I don't see why any of the following wouldn't be possible:
> 
>   1. When using --save-temps, use cpp output routines to write the .i
>      file while passing the tokens directly to cc1.
> 
>   2. Invoke cpp to generate the .i file, then do it again for
>      compilation as usual.

The problem with these two is that the primary use of --save-temps is
to generate a .i file which the maintainers can feed into cc1 to show
a bug.  With the current arrangement, if the user sees a bug with
--save-temps then the maintainers should see it too because they are
running cc1 exactly the same way.  With either of these, the
maintainers will be running cc1 differently to the user and the bug
might mutate.

>   2. Position tokens in the .i file such that they are at the correct
>      column numbers.  This would be useful for other programs too.

This would probably cause the .i file size to explode.

-- 
- Geoffrey Keating <geoffk@geoffk.org>

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]