emacs.c broken by revision 1.435

emacs.c broken by revision 1.435

[Tim Van Holder]

The "LINUX_SBRK_BUG" condition was incorrectly replaced with

#if (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
  __sbrk (1);
#endif

instead of

#if defined (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
  __sbrk (1);
#endif

resulting in a broken build.

I'm also not quite certain why a condition that was only present in an
Alpha header is now being used for all platforms. But I suppose the
__sbrk call is sufficiently harmless?

Re: emacs.c broken by revision 1.435

[Miles Bader]

"Tim Van Holder" <tim.vanholder@gmail.com> writes:
> The "LINUX_SBRK_BUG" condition was incorrectly replaced with
>
> #if (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>
> instead of
>
> #if defined (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>
> resulting in a broken build.

Why does the first result in a broken build?  Seems to work fine with
the C compilers I tried it with... (both with GNU_LINUX defined, and
with GNU_LINUX undefined)

-Miles

-- 
`Suppose Korea goes to the World Cup final against Japan and wins,' Moon said.
`All the past could be forgiven.'   [NYT]

Re: emacs.c broken by revision 1.435

[Yavor Doganov]

Miles Bader wrote:
> "Tim Van Holder" <tim.vanholder@gmail.com> writes:
> > The "LINUX_SBRK_BUG" condition was incorrectly replaced with
> >
> > #if (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>
> Why does the first result in a broken build?

With GCC 4.3.1 on GNU/Linux:

emacs.c:830:15: error: missing expression between '(' and ')'

Re: emacs.c broken by revision 1.435

[David Kastrup]

"Tim Van Holder" <tim.vanholder@gmail.com> writes:

> The "LINUX_SBRK_BUG" condition was incorrectly replaced with
>
> #if (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>   __sbrk (1);
> #endif
>
> instead of
>
> #if defined (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>   __sbrk (1);
> #endif
>
> resulting in a broken build.

Can we rely on "defined" being available?  Seems like an ANSI feature to
me...  Maybe nested #ifdef and #if would be safer?

-- 
David Kastrup

Re: emacs.c broken by revision 1.435

[Miles Bader]

David Kastrup <dak@gnu.org> writes:
> Can we rely on "defined" being available?  Seems like an ANSI feature to
> me...  Maybe nested #ifdef and #if would be safer?

There's no way in hell that Emacs is going to compile with a pre-ansi
compiler, and trying to change that would be an absurd waste of time.

-Miles

-- 
Run away!  Run away!

Re: emacs.c broken by revision 1.435

[Juanma Barranquero]

On Wed, Jul 23, 2008 at 10:48, David Kastrup <dak@gnu.org> wrote:

> Can we rely on "defined" being available?

??

There are at least 490 instances of "#if defined" and "#if !defined"
in the src/ tree.

  Juanma

Re: emacs.c broken by revision 1.435

[David Kastrup]

"Juanma Barranquero" <lekktu@gmail.com> writes:

> On Wed, Jul 23, 2008 at 10:48, David Kastrup <dak@gnu.org> wrote:
>
>> Can we rely on "defined" being available?
>
> ??
>
> There are at least 490 instances of "#if defined" and "#if !defined"
> in the src/ tree.

I stand admonished.

-- 
David Kastrup

Re: emacs.c broken by revision 1.435

[Juanma Barranquero]

On Wed, Jul 23, 2008 at 11:21, David Kastrup <dak@gnu.org> wrote:

> I stand admonished.

Not my intention at all. I was just surprised that you hadn't noticed
the "defined"s before (because they're quite conspicuous).

  Juanma

Re: emacs.c broken by revision 1.435

[Jason Rumney]

David Kastrup wrote:

> Can we rely on "defined" being available?  Seems like an ANSI feature to
> me...  Maybe nested #ifdef and #if would be safer?

I think we already rely on it being available to such an extent that we
would have caught any problems with its use by now.

Re: emacs.c broken by revision 1.435

[Stefan Monnier]

> The "LINUX_SBRK_BUG" condition was incorrectly replaced with
> #if (GNU_LINUX) && __GNU_LIBRARY__ - 0 < 6
>   __sbrk (1);
> #endif

Indeed Dan, this looks wrong.  Beside the obvious lack of "defined",
this change activates this code for all platforms rather than just
Alpha, and it makes the code less readable: the previous use of the
"LINUX_SBRK_BUG" identifier made it clear what the purpose was, and
clearly separated the bug&bugfix part from the detection of the bug.
I.e. the previous code was better.


        Stefan

Re: emacs.c broken by revision 1.435

[Dan Nicolaescu]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

  > I.e. the previous code was better.

The change was undone.