Move SSA macros and declarations out of tree.h

[Basile Starynkevitch]

[slightly off topic about the specific patch discussed, but on topic
with the comments]

On Fri, Oct 01, 2010 at 03:00:06PM +0200, Manuel López-Ibáñez wrote:
> On 1 October 2010 13:57, Richard Guenther <richard.guenther@gmail.com> wrote:
> >
> > Note that IMHO our headers should mainly follow the implementation,
> > not serve as high-level documentation.
> 
> It has been stated several times that the sources are the only
> up-to-date documentation and there are no hopes of keeping an
> up-to-date documentation (since maintainers do not enforce this on
> themselves).

We might find that state of affairs pitiful. 

Let's take as an example the documentation of our nearly 250
passes. So far, it is very incomplete
http://gcc.gnu.org/onlinedocs/gccint/Passes.html most gcc passes are
not individually documented in gcc/doc/passes.texi

We could imagine that every struct opt_pass (or some "sub-class" of it
like struct rtl_opt_pass, struct gimple_opt_pass, ...) would have
either a specific comment (or perhaps an additional string field)
which is a short one sentence description.  Then we could have a small
(perhaps awk) script which extract all these sentences to make a
texinfo table of all documented passes.  That would put a much smaller
burden on developpers than expecting each of them to add a paragraph
to gcc/doc/passes.texi (which did not happen in the past and probably
won't happen in the future). But going that [easier] route means that
we accept to derive a piece of [*.texi] documentation from a piece of
[*.c] code.

FWIW, in GCC MELT I had a similar issue. See
http://gcc.gnu.org/ml/gcc/2010-05/msg00539.html and many other
messages.

Regards.

-- 
Basile STARYNKEVITCH         http://starynkevitch.net/Basile/
email: basile<at>starynkevitch<dot>net mobile: +33 6 8501 2359
8, rue de la Faiencerie, 92340 Bourg La Reine, France
*** opinions {are only mines, sont seulement les miennes} ***