Re: [PATCH] Update backward/algo.h

[Paolo Carlini <pcarlini at unitus dot it>]

Nathan Myers wrote:

> On Sun, Dec 30, 2001 at 09:07:22AM +0100, Paolo Carlini wrote:
> > this, rather harmless patch IMHO, updates backward/algo.h in the
> > light of the recent changes to include/bits/stl_algo.h and the new
> > include/ext/algorithm.  Tested i686-pc-linux-gnu.
> >
> > *** libstdc++-v3-orig/include/backward/algo.h Wed Jun 27 19:09:52 2001
> > --- libstdc++-v3/include/backward/algo.h Sun Dec 30 09:00:49 2001
> > ***************
> > *** 62,67 ****
> > --- 62,68 ----
> >   #include "iterator.h"
> >   #include <bits/stl_algo.h>
> >   #include <bits/stl_numeric.h>
> > + #include <ext/algorithm>
>
> Unless I'm confused (which is quite possible), this violates the
> in-library header inclusion rule that forbids any header from
> sub-including another header whose name has no suffix.
>
>   http://gcc.gnu.org/onlinedocs/libstdc++/17_intro/HEADER_POLICY
>
> I wrote the policy back in '98.  As noted, much of the motivation
> for the rule was GNU Make's (and others') eagerness to delete files
> that have no suffix, in the event of a fatal signal.

I see. I was not aware of all of this... :-(

> It may be necessary to give ext/algorithm a suffix,

This I would rather prefer not, since ext/algorithm can be directly included by
the user, and it would be the *only* header directly included by the user (not
considering the backward ones, of course) having a suffix .h. Do you agree?

> or to make it
> sub-include ext/bits/algorithm.h, and make algo.h include that
> directly.

My preference goes to this option. Yours?
But first I would like to post (just a couple of hours) my patch moving all the
estensions already present inside ext (rope, slist, hash_map, hash_set) to
__gnu_cxx.

Thanks for your feedback,
Paolo.