This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] add alloc_size attribute
- From: "Attila Feher F \(JO/LMF\)" <attila dot f dot feher at ericsson dot com>
- To: <gcc-patches at gcc dot gnu dot org>
- Cc: <dmueller at suse dot de>, <meissner at suse dot de>, <richard dot guenther at gmail dot com>, <wolof at freemail dot hu>
- Date: Wed, 25 Apr 2007 16:12:59 +0200
- Subject: Re: [PATCH] add alloc_size attribute
Hi All,
Benjamin Kosnik forwarded this to me as he knows I am working on
standardizing alignment specifications and other alignment related
things (eg.: alingof).
I see that you plan to provide alignment support into the memory
allocation system of C and C++. This is what I wrote to Benjamin:
8<---
*But* the point is that POSIX and WG21 will be discussing adding
alignment support for memory allocators (eg.: posix_memalign) in Toronto
this summer and that will certainly be followed by (if there will be
someone making a paper) extending the C++ allocator(s) (operator new
variant) with the same feature. It would be *very* unfortunate and
damaging if gcc/g++ would preempt this standardization by introducing a
non-standardized feature (that has no industry wide consensus behind it)
in the last minute. Let's do this right. This extension affects the
core language as well as the library on a way that an implementation
cannot support different standards. So if what gcc chooses does not
satisfy the rest of the industry gcc will either loose standard
compliance or force its users to change their code. I believe that the
time we can get this done the right way is not so much that we should
take the risky road.
--->8
What I am trying to say is: let's work together and get a standard
solution done in the first place and that I am ready to help with this
effort. All I need is info on what are the driving use cases, what is
the suggested solution, how does it effect compatibility etc. and we can
put a paper together and get the thoughts of all other
vendors/implementers.
Since g++ has a very very large user base I believe it is exceptionally
important that we try to ensure that the way this feature is implemented
in g++ can be supported by everyone. And thereby can be part of the
standard. I believe this approach will benefit everyone, including the
creators and the users.
I am ready to try to organize this effort and get alignment support into
the allocators *if* I get help from those who have the use cases.
Please feel free to contact me at wolof at freemail hu and we can
prepare a paper. If we are fast enough we can get into the post-Oxford
mailing (May 5 IIRC).
Please make sure that you CC answers to the wolof freemail hu address so
that I will get them. I am not on the mailing list (excuses:
overwhelmed, I'm rather a user so I wouldn't understand most of the
stuff etc. ;-)
I know I am probably completely unknown to the g++ community and that it
seems a bit out of place for me to ask you to hold back you patches.
But believe me, all what drives me is to help, I have no hidden agenda.
Neither I am a bureaucrat - I am a happy user of g++. :-)
TIA,
Attila