GCC 4.9.1 suggests -std=c99

GCC 4.9.1 suggests -std=c99

[Dmitry Antipov]

With GCC 4.9.1 and '--enable-checking --enable-check-lisp-object-type --enable-gcc-warnings':

../../trunk/src/sysdep.c: In function 'str_collate':
../../trunk/src/sysdep.c:3751:2: error: 'for' loop initial declarations are only allowed in C99 or C11 mode
   for (int i = 1; i < 3; i++)
   ^
../../trunk/src/sysdep.c:3751:2: note: use option -std=c99, -std=gnu99, -std=c11 or -std=gnu11 to compile your code
../../trunk/src/sysdep.c:3766:2: error: 'for' loop initial declarations are only allowed in C99 or C11 mode
   for (int i = 1; i < 3; i++)
   ^
I.e. shouldn't we always use -std=c99 if we assume C99 anyway?

Dmitry

Re: GCC 4.9.1 suggests -std=c99

[Eli Zaretskii]

> Date: Tue, 09 Sep 2014 18:30:06 +0400
> From: Dmitry Antipov <dmantipov@yandex.ru>
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> I.e. shouldn't we always use -std=c99 if we assume C99 anyway?

Not -std=c99, in any case.  Perhaps -std=gnu99.  We don't want strict
adherence to ANSI C99, it will flag too many false problems.

Re: GCC 4.9.1 suggests -std=c99

[Paul Eggert]

Dmitry Antipov wrote:
> With GCC 4.9.1 and '--enable-checking --enable-check-lisp-object-type
> --enable-gcc-warnings':
>
> ../../trunk/src/sysdep.c: In function 'str_collate':
> ../../trunk/src/sysdep.c:3751:2: error: 'for' loop initial declarations
> are only allowed in C99 or C11 mode

I don't observe this problem when I build with GCC 4.9.1.  src/Makefile 
says "CC = gcc -std=gnu99", so it works.  Perhaps you overrode CC?  Or 
maybe you have an old version of autoconf?

While we're on the topic, --enable-gcc-warnings generates false alarms 
when combined with arbitrary other options like 
--enable-check-lisp-object-type, CFLAGS='-g3 -O0', etc.  I do not take 
the trouble to pacify GCC for all combinations of build options, just 
for the default ones.

Re: GCC 4.9.1 suggests -std=c99

[Eli Zaretskii]

> Date: Tue, 09 Sep 2014 07:58:54 -0700
> From: Paul Eggert <eggert@cs.ucla.edu>
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> Dmitry Antipov wrote:
> > With GCC 4.9.1 and '--enable-checking --enable-check-lisp-object-type
> > --enable-gcc-warnings':
> >
> > ../../trunk/src/sysdep.c: In function 'str_collate':
> > ../../trunk/src/sysdep.c:3751:2: error: 'for' loop initial declarations
> > are only allowed in C99 or C11 mode
> 
> I don't observe this problem when I build with GCC 4.9.1.

But do we actually want to use such initializations inside a for loop?

Re: GCC 4.9.1 suggests -std=c99

[Paul Eggert]

Eli Zaretskii wrote:
> do we actually want to use such initializations inside a for loop?

It's a better style, as it scopes the control variable better.  And we 
are assuming C99 in the trunk nowadays.

It will be a portability problem only if we run into a pre-C99 compiler 
that supports C99-style statements-before-declarations (which we're 
already using in the trunk) but does not support a C99-style declaration 
as the first clause of a for-loop.  I'd be a bit surprised if such a 
compiler were in current use (but then I've been surprised before ...).

Re: GCC 4.9.1 suggests -std=c99

[Dmitry Antipov]

On 09/09/2014 06:58 PM, Paul Eggert wrote:

> I don't observe this problem when I build with GCC 4.9.1.  src/Makefile says
> "CC = gcc -std=gnu99", so it works.  Perhaps you overrode CC?

Argh, yes, it was an explicit CC=... in make.

Dmitry

Re: GCC 4.9.1 suggests -std=c99

[Eli Zaretskii]

> Date: Tue, 09 Sep 2014 08:21:54 -0700
> From: Paul Eggert <eggert@cs.ucla.edu>
> CC: dmantipov@yandex.ru, emacs-devel@gnu.org
> 
> Eli Zaretskii wrote:
> > do we actually want to use such initializations inside a for loop?
> 
> It's a better style, as it scopes the control variable better.  And we 
> are assuming C99 in the trunk nowadays.
> 
> It will be a portability problem only if we run into a pre-C99 compiler 
> that supports C99-style statements-before-declarations (which we're 
> already using in the trunk) but does not support a C99-style declaration 
> as the first clause of a for-loop.  I'd be a bit surprised if such a 
> compiler were in current use (but then I've been surprised before ...).

I think we should have discussed whether to use these C99 features
before actually starting using them.  Emacs definitely has certain
coding style, which doesn't include these.

I understand the rationale for using C99 features for which there's no
reasonably practical alternatives, or rely on C99 semantics where
earlier versions left things undefined.  But the above don't belong to
either class.

Re: GCC 4.9.1 suggests -std=c99

[Dmitry Antipov]

On 09/09/2014 07:50 PM, Eli Zaretskii wrote:

>> It's a better style

IMHO only if it is just a control variable, and in the only loop.
If it isn't, e.g. in:

int i;
...
for (i = 0; i < XXX; i++)
  ...
i = a;
a = b;
b = i;
   ...
for (i = 0; i < YYY; i++)
   ...

then C99/C++-style:

...
for (int i = 0; i < XXX; i++)
  ...
int i = a;
a = b;
b = i;
   ...
for (int i = 0; i < YYY; i++)
   ...

becomes a monkey's typing exercise.

Dmitry

Re: GCC 4.9.1 suggests -std=c99

[Paul Eggert]

Eli Zaretskii wrote:
> we should have discussed whether to use these C99 features
> before actually starting using them.

We discussed them in Bug#17487.  As I mentioned, Emacs developers 
routinely made changes that used these features, I was occasionally 
auditing the code and removing their use for no practical reason (only 
because of standards-conformance issues), and I got tired of this 
pointless task.

My guess is that when Michael Albinus introduced the 'for (int i = 1; 
...' loop in trunk bzr 117771, he wasn't consciously thinking "Oh boy! I 
get to use C99 now! I will now change Emacs's style!"  Most likely he 
was just using a common and natural idiom that worked for him, and 
didn't notice that it wouldn't port to C89 and was therefore contrary to 
Emacs's old style rules.

> Emacs definitely has certain
> coding style, which doesn't include these.

The part of Emacs's style that disallowed these C99 features was there 
primarily because of C89's restrictions.  It's often not the best style 
and we should not insist on it.

I agree with Dmitry that we should not insist on a declaration as the 
first clause of every for loop, but there doesn't seem to be a good 
reason to prohibit a declaration there, whereas there is a good reason 
(more-accurate scoping) to allow it.