Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stefan Monnier]

>   * src/dbusbind.c: Use #include <config.h> instead of "config.h" as all
>   other files do.

I understand teh desire to make it uniform, but why choose to
standardize on <config.h> rather than "config.h"?
I'd have thought the other one would be a better choice, since <...> is
normally used for system headers and "..." for the program's
own headers.


        Stefan

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stephen J. Turnbull]

Stefan Monnier writes:
 > >   * src/dbusbind.c: Use #include <config.h> instead of "config.h" as all
 > >   other files do.
 > 
 > I understand teh desire to make it uniform, but why choose to
 > standardize on <config.h> rather than "config.h"?

VPATH.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stefan Monnier]

>> >   * src/dbusbind.c: Use #include <config.h> instead of "config.h" as all
>> >   other files do.
>> I understand teh desire to make it uniform, but why choose to
>> standardize on <config.h> rather than "config.h"?
> VPATH.

I don't see how that relates.  VPATH seems to be a make-only feature,
whereas we're talking about C code.


        Stefan

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Dan Nicolaescu]

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

>>   * src/dbusbind.c: Use #include <config.h> instead of "config.h" as all
>>   other files do.
>
> I understand teh desire to make it uniform, but why choose to
> standardize on <config.h> rather than "config.h"?

Just because <config.h> was used in more places and for a longer time.
My own preference would be "config.h", but it looks like the opposite
choice was made a while ago...

> I'd have thought the other one would be a better choice, since <...> is
> normally used for system headers and "..." for the program's
> own headers.
>
>
>         Stefan

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stephen J. Turnbull]

Stefan Monnier writes:
 > >> >   * src/dbusbind.c: Use #include <config.h> instead of "config.h" as all
 > >> >   other files do.
 > >> I understand teh desire to make it uniform, but why choose to
 > >> standardize on <config.h> rather than "config.h"?
 > > VPATH.
 > 
 > I don't see how that relates.  VPATH seems to be a make-only feature,
 > whereas we're talking about C code.

Specifically, people who build both in-tree and out-of-tree.

If you do that, VPATH/--srcdir can break because you have multiple
config.h on the #include search path, and the one in-tree always wins
if you use #include "config.h".

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Jan Djärv]

Stephen J. Turnbull skrev 2010-06-02 18.55:
> Stefan Monnier writes:
>   >  >>  >    * src/dbusbind.c: Use #include<config.h>  instead of "config.h" as all
>   >  >>  >    other files do.
>   >  >>  I understand teh desire to make it uniform, but why choose to
>   >  >>  standardize on<config.h>  rather than "config.h"?
>   >  >  VPATH.
>   >
>   >  I don't see how that relates.  VPATH seems to be a make-only feature,
>   >  whereas we're talking about C code.
>
> Specifically, people who build both in-tree and out-of-tree.
>
> If you do that, VPATH/--srcdir can break because you have multiple
> config.h on the #include search path, and the one in-tree always wins
> if you use #include "config.h".

I get that once in a while, but if you are building out of the tree you 
shouldn't have any config.h in-tree.  Unless you mix in and out of tree builds.

	Jan D.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stephen J. Turnbull]

Jan Djärv writes:

 > I get that once in a while, but if you are building out of the tree
 > you shouldn't have any config.h in-tree.  Unless you mix in and out
 > of tree builds.
 
Personally, I use out-of-tree builds exclusively.  XEmacs consistently
uses '#include <config.h>' because somebody did mix builds, though.

I suspect it's actually a common pattern among new developers (and
Bazaar will contribute to continuing temptation because it's so slow)
to do

tar ...
cd ...
./configure -Oinfinity ...
make
# didn't see *that* coming
mkdir debug-build
cd debug-build
../configure --srcdir=.. -O $DEBUG_OPTIONS ...
make

rather than make a new workspace.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Miles Bader]

"Stephen J. Turnbull" <stephen@xemacs.org> writes:
> Specifically, people who build both in-tree and out-of-tree.

Does anybody actually do that?  It's a very bad habit to get into,
because it often doesn't work.

-Miles

-- 
I'd rather be consing.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Stephen J. Turnbull]

Miles Bader writes:
 > "Stephen J. Turnbull" <stephen@xemacs.org> writes:
 > > Specifically, people who build both in-tree and out-of-tree.
 > 
 > Does anybody actually do that?  It's a very bad habit to get into,
 > because it often doesn't work.

I don't know if anybody does it habitually; if they do, problems
haven't been reported where I can see them.  I've never had problems
when I do it (not very often, usually by accident or greater than
usual laziness).  The exception is when a new file gets an '#include
"config.h"', but that's a bug.

Re: [Emacs-diffs] /srv/bzr/emacs/trunk r100503: Fix config.h includes.

[Richard Stallman]

    Just because <config.h> was used in more places and for a longer time.
    My own preference would be "config.h", but it looks like the opposite
    choice was made a while ago...

There was a reason for this choice.  I think it had to do with use of
the -I option in certain cases.