Fwd: -Wpointer-sign for GCC 4.1

Fwd: -Wpointer-sign for GCC 4.1

[Jim Blandy]

The message below is kind of odd.  We do use -Wall, so if the pointer
sign warning will be printed when -Wall is specified, we'll still need
to pass an explicit argument to disable it.  Which doesn't exactly
take the decision out of our hands, as we were hoping.

I think we should decide, for ourselves, whether we think the warning
is helpful or not, and then not be demure about doing the necessary
GCC stuff to enable or disable it.  Hoping GCC would answer the
question for us was dopey.

I think there's some documentation value in reserving gdb_byte for
binary blobs and char for host-format text.  It wouldn't have been
worth it before, but at this point we've got fixes for almost all
those warnings in place; we can't get those hours back, so the
cost/benefit is different now.  So I think we should continue to
request the warning.

---------- Forwarded message ----------
From: Joe Buck <Joe.Buck@synopsys.com>
Date: Jan 18, 2006 9:31 AM
Subject: Re: -Wpointer-sign for GCC 4.1
To: Mike Stump <mrs@apple.com>
Cc: Daniel Jacobowitz <drow@false.org>, gcc@gcc.gnu.org


On Wed, Jan 18, 2006 at 01:10:19AM -0800, Mike Stump wrote:
> On Jan 17, 2006, at 1:19 PM, Daniel Jacobowitz wrote:
> >Someone's informed Richard Stallman that this (annoying) warning
> >will not be
> >enabled by default in GCC 4.1.
>
> >But, it currently seems to be.  Should it be turned off before the
> >release?
>
> The SC or Jim Wilson will know more than I.
>
> >  If not, who told RMS it was? :-)
>
> Likewise.

The Emacs developers were unhappy, so RMS complained to the SC, and there
was a discussion.  At first RMS just wanted the warning to appear only
with --pedantic or -Wpointer-sign, but he was convinced that it should
also appear with -Wall.

So the answer is that, after consulting with some of the affected people
(which is why you got mail, Mike) the SC told RMS that it would be
changed.  If it hasn't been done yet, then it's a release blocker,
because it was a promise the SC made.

Re: Fwd: -Wpointer-sign for GCC 4.1

[Daniel Jacobowitz]

On Wed, Jan 18, 2006 at 10:40:55AM -0800, Jim Blandy wrote:
> The message below is kind of odd.  We do use -Wall

No, we don't.  See the comments in configure.ac about this.

>, so if the pointer
> sign warning will be printed when -Wall is specified, we'll still need
> to pass an explicit argument to disable it.  Which doesn't exactly
> take the decision out of our hands, as we were hoping.
> 
> I think we should decide, for ourselves, whether we think the warning
> is helpful or not, and then not be demure about doing the necessary
> GCC stuff to enable or disable it.  Hoping GCC would answer the
> question for us was dopey.

Sorry, that wasn't my goal.  My goal with forwarding this to the GCC
list was independent of what GDB should do - as Joe wrote, the GCC SC
made a promise to RMS that I'm trying to make sure doesn't slip through
the cracks.

> I think there's some documentation value in reserving gdb_byte for
> binary blobs and char for host-format text.  It wouldn't have been
> worth it before, but at this point we've got fixes for almost all
> those warnings in place; we can't get those hours back, so the
> cost/benefit is different now.  So I think we should continue to
> request the warning.

Almost all of those warnings on some targets, btw.  An ARM cross
debugger has a whole new set of them.  I don't know which option I
prefer.  Either way I'll probably clean up ARM soonish.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Fwd: -Wpointer-sign for GCC 4.1

[Jim Blandy]

On 1/18/06, Daniel Jacobowitz <drow@false.org> wrote:
> On Wed, Jan 18, 2006 at 10:40:55AM -0800, Jim Blandy wrote:
> > The message below is kind of odd.  We do use -Wall
>
> No, we don't.  See the comments in configure.ac about this.

D'oh.  You're right, of course.

> Sorry, that wasn't my goal.  My goal with forwarding this to the GCC
> list was independent of what GDB should do - as Joe wrote, the GCC SC
> made a promise to RMS that I'm trying to make sure doesn't slip through
> the cracks.

Okay.  I'd gotten the impression from your conversation with Eli that
the plan was to just forget the whole thing if future GCC's were going
to drop the warning.

Re: Fwd: -Wpointer-sign for GCC 4.1

[Daniel Jacobowitz]

On Wed, Jan 18, 2006 at 10:58:59AM -0800, Jim Blandy wrote:
> On 1/18/06, Daniel Jacobowitz <drow@false.org> wrote:
> > On Wed, Jan 18, 2006 at 10:40:55AM -0800, Jim Blandy wrote:
> > > The message below is kind of odd.  We do use -Wall
> >
> > No, we don't.  See the comments in configure.ac about this.
> 
> D'oh.  You're right, of course.
> 
> > Sorry, that wasn't my goal.  My goal with forwarding this to the GCC
> > list was independent of what GDB should do - as Joe wrote, the GCC SC
> > made a promise to RMS that I'm trying to make sure doesn't slip through
> > the cracks.
> 
> Okay.  I'd gotten the impression from your conversation with Eli that
> the plan was to just forget the whole thing if future GCC's were going
> to drop the warning.

Nah - my plan is to forget the whole thing if we decide to, but no one
wants to decide.

You and I and a couple of others in the past have squashed most of
these warnings in GDB.  At the moment I'm not convinced that running
around hammering on the rest of them is a worthwhile use of our time,
though.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Fwd: -Wpointer-sign for GCC 4.1

[Eli Zaretskii]

> Date: Wed, 18 Jan 2006 14:01:20 -0500
> From: Daniel Jacobowitz <drow@false.org>
> 
> > Okay.  I'd gotten the impression from your conversation with Eli that
> > the plan was to just forget the whole thing if future GCC's were going
> > to drop the warning.
> 
> Nah - my plan is to forget the whole thing if we decide to, but no one
> wants to decide.

I don't know about ``no one'', I think I already said that I
considered the rest of these warnings useless annoyances.

> At the moment I'm not convinced that running around hammering on the
> rest of them is a worthwhile use of our time, though.

FWIW, I think it isn't a worthwhile use of our time.

Re: Fwd: -Wpointer-sign for GCC 4.1

[Mark Kettenis]

> Date: Wed, 18 Jan 2006 14:01:20 -0500
> From: Daniel Jacobowitz <drow@false.org>
> 
> On Wed, Jan 18, 2006 at 10:58:59AM -0800, Jim Blandy wrote:
> > On 1/18/06, Daniel Jacobowitz <drow@false.org> wrote:
> > > On Wed, Jan 18, 2006 at 10:40:55AM -0800, Jim Blandy wrote:
> > > > The message below is kind of odd.  We do use -Wall
> > >
> > > No, we don't.  See the comments in configure.ac about this.
> > 
> > D'oh.  You're right, of course.
> > 
> > > Sorry, that wasn't my goal.  My goal with forwarding this to the GCC
> > > list was independent of what GDB should do - as Joe wrote, the GCC SC
> > > made a promise to RMS that I'm trying to make sure doesn't slip through
> > > the cracks.
> > 
> > Okay.  I'd gotten the impression from your conversation with Eli that
> > the plan was to just forget the whole thing if future GCC's were going
> > to drop the warning.
> 
> Nah - my plan is to forget the whole thing if we decide to, but no one
> wants to decide.

Let's forget about them.  The mostly bogus signed vs. unsigned char
ones dominate the useful warnings for the other integers.

Is there any reason not to commit my -Werror patch now (with the
-Wno-pointer-sign)?

Mark

Re: Fwd: -Wpointer-sign for GCC 4.1

[Daniel Jacobowitz]

On Wed, Jan 18, 2006 at 11:24:09PM +0100, Mark Kettenis wrote:
> Let's forget about them.  The mostly bogus signed vs. unsigned char
> ones dominate the useful warnings for the other integers.
> 
> Is there any reason not to commit my -Werror patch now (with the
> -Wno-pointer-sign)?

It's fine with me, but I'd like to make sure Jim agrees first; I'm not
sure if he's said or not.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Fwd: -Wpointer-sign for GCC 4.1

[Jim Blandy]

On 1/18/06, Daniel Jacobowitz <drow@false.org> wrote:
> Nah - my plan is to forget the whole thing if we decide to, but no one
> wants to decide.

I think we should encourage the use of gdb_byte for binary blobs in
new code (say, at patch review time).  But it sounds like you estimate
there's still a ways to go to get target-specific code in line.  Since
this concurrence between you and Eli:

>> Shall we discard the remaining (more minor) cases and turn off this
>> warning?
>
>I think so, yes.  We have better uses for our energy right now.

wasn't conditioned on GCC's future choices, I'm happy to join it.

Re: Fwd: -Wpointer-sign for GCC 4.1

[Eli Zaretskii]

> Date: Wed, 18 Jan 2006 16:50:42 -0800
> From: Jim Blandy <jimb@red-bean.com>
> 
> I think we should encourage the use of gdb_byte for binary blobs in
> new code (say, at patch review time).

100% agreement.