This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: New ARI warning Wed May 23 01:55:03 UTC 2012
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: brobecker at adacore dot com
- Cc: tromey at redhat dot com, mark dot kettenis at xs4all dot nl, dje at google dot com, pierre dot muller at ics-cnrs dot unistra dot fr, gdb-patches at sourceware dot org
- Date: Tue, 26 Jun 2012 15:30:10 +0200 (CEST)
- Subject: Re: New ARI warning Wed May 23 01:55:03 UTC 2012
- References: <CADPb22QBYSRjoqrgX7p0d5KdAZ+kEk4Ga2YrfHMrm_t_yKCE6w@mail.gmail.com> <CADPb22QOSCV240MV2m0KmsXfNb9O5_WsA+7Jsjn1UJwigaNTrQ@mail.gmail.com> <4fbc9d77.0853b40a.641e.ffff90dbSMTPIN_ADDED@mx.google.com> <CADPb22Qi0Zda439aieEGu26+GWf0y8XzBUkcpg-Kw1BUcjvPNA@mail.gmail.com> <87bold8l4d.fsf@fleche.redhat.com> <201205282043.q4SKhksB010254@glazunov.sibelius.xs4all.nl> <87ipejib8o.fsf@fleche.redhat.com> <20120622171922.GK2799@adacore.com> <20120622173049.GL2799@adacore.com> <877guzgs7r.fsf@fleche.redhat.com> <20120622190228.GM2799@adacore.com>
> Date: Fri, 22 Jun 2012 12:02:28 -0700
> From: Joel Brobecker <brobecker@adacore.com>
>
> > >> * configure.ac (build_warnings): Add -Wdeclaration-after-statement.
> > >> * configure: Regenerate.
> >
> > Yes, ok. Thanks; I hadn't realized there was a flag for this already :)
>
> Thanks, checked in.
>
> > Discussion on irc pointed out that this is still allowed:
> >
> > for (int i = 0; ...)
> >
> > I think this doesn't suffer from the readability problems that
> > declarations in the code generally do; and in fact usually makes the
> > code cleaner, by restricting the scope of the loop variable.
>
> Agreed.
>
> Note that this is only going to be accepted if we compile in
> C99 mode, I think. Otherwise, you'll get a warning which is
> unrelated declarations being used after statements.
>
> error: 'for' loop initial declarations are only allowed in C99 mode
>
> > How about we flip the switch to C99 for 7.6?
>
> Sounds good to me. Do we want to be exclusive, rather than inclusive?
> In other words, say: The following C99 constructs are allowed, and
> maintain that list, rather that allow all of C99, and then list
> the features not allowed. I understand that some features are still
> not implemented (or portable?).
I'm not aware of any C99 language features that aren't widely
implemented by the halfway modern compilers. But there might be some
obscure corners of the standard that I'm not familiar with. I think
only the approach of listing features that we don't allow makes sense.
That list should be very short.
C99 library functions and header files are a different matter though.
Those may not be available on older systems. We can only use those if
we provide proper replacement functions (e.g. gnulib).