This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: gengtype improvements for plugins, completed! patch 1/N [declprog]
On Thu, Sep 09, 2010 at 10:23:15AM +0300, Laurynas Biveinis wrote:
> 2010/9/9 Basile Starynkevitch <basile@starynkevitch.net>:
> > On Thu, 9 Sep 2010 09:35:55 +0300
>
> > In particular, I find strong the requirement of explaining
> > any temporary fix which will be dismantled by the next patch. And such
> > fixes are implied by the requirement of making gengtype still work as
> > before at every piece of the patch set.
>
> Oh. I might be wrong, but I don't think that's necessary. You have a
> set of patches which are interdependent. I think it is fine for an
> arbitrary subset of these patches to break something as long as the
> end result is OK. Can you point to where you were told to make sure
> that gengtype works as before _at every step_?
In http://gcc.gnu.org/ml/gcc/2010-08/msg00353.html Ian Taylor wrote in
a reply to my questions
>> An ideal sequence of patches is a series of patches which may be applied
>> in order. At each stage the tree should be buildable. The patches
>> don't have to be independent--they can be applied in a specific
>> order--but they should each work.
I understood the above paragraph from Ian as requirement that gengtype
should always work at every stage, i.e. _at every step_ to quote
Laurynas.
Maybe I misunderstood, or took as a requirement what is just an ideal
wish, that is something which is not realistically achievable but just
wanted in a Platonic world which does not exist in GCC.
And believe me, gengtype, even with our patch set, is *very far* from
being ideal. It will remain crappy forever
because garbage collection is a *global* property of a program, and
there is still no consensus within the GCC community about it; my
"purist" view is that the GCC community should require every
dynamically allocated data to be garbage collected with almost no
exceptions, and then implement the tools [GC support à la Ggc but
better, code generators much better than gengtype] to make that
happen. But I do know that most people disagree with my disruptive
& extreme position. Notice that transition to C++ could help and be
an incentive for better garbage collection (C++ permit virtual
functions as marking routines and could wrap nicely local GC-ed
pointers). But in general, as long as there is no strong consensus
on any *global* property of GCC, we won't be able to implement it
cleanly and robustly. This is true of every *global* property or
feature of GCC, e.g. memory management, organization of passes, code
formatting style, or GCC documentation.
Cheers.
--
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} ***