gcc newsletter #11

[Mathieu Lacage <Mathieu dot Lacage at sophia dot inria dot fr>]

hi all,

Have some fun,

http://gccnews.chatta.us

Inline:

gcc release

I think I forgot to mention that gcc mainline (that is, what will become
3.5) is now in stage 3[0] which means that no new features will go in
except those approved in the 3.5 plan[1] layed out by Mark. Of course,
Mark sent a timely release status[2]. Obviously, memory usage[3] still
needs to be improved.

Gabriel Dos Reis released a 3.3.5 prerelease[4] for testing.

As a side note, it has now been officially[5] decided that the next
version of gcc would be 4.0.0. This ends many past lengthy
threads/discussions. On a related front, Nathan Sidwell was appointed
co-maintainer of the C++ frontend together with Mark Mitchell and Jason
Merrill.

gcc development

Daniel Berlin did setup a gcc wiki which is now running on
dberlin.org[6]. He started adding content there which means you are
encouraged to check his work regularly for updates. Of course, this
effort spurted a discussion[7] on what the goal of this wiki is.

Ranjit Mathew pointed me to this patch[8] by Richard Henderson which
implements stack slot coalescing. It was also pointed to me by Kaveh R.
Ghazi that the sentinel patch presented here last week got in[9].

The Ada frontend has seen recently a lot of criticism from the community
and the other gcc contributors. Currently, this frontend is not part of
the release criterion for the next gcc release because it was felt that
ACT[10] (the sole company which provides commercial support for this Ada
frontend and where most if not all the Ada contributors work) would not
be able to complete in time a correct port of the frontend to the
tree-ssa architecture. Much to the surprise of many developers, Richard
Kenner seems to have almost completed the port and work is underway to
merge these changes to the CVS repository.

However, the Ada maintainers have long suffered (benefited ?) from a
special agreement with the gcc developers that their patches would not
need to follow the normal gcc development process. Namely, individual
public patch review is not necessary. Public Testcases are not necessary
to fix a bug. Of course, it is clear to everyone involved that ACT does
perform careful patch reviews internally and owns a huge testcase
database but the crux of the problem is that none of these are shared
publicly. The gcc developers all seem to agree that this indulgence
should be no more which means that patches should be broken down into
individual patches and reviewed publicly. Patch batching[11] is to die.
Not surprisingly, they are finding a hard time convincing the Ada
developers. Zack Weinberg summarizes[12] well the general concensus and
insists[13] despite the answers by Robert Dewar who worries about the
proprietary software of his customers.

Discussions surrounding this issue have constituted the bulk of the
mailing list traffic recently. Richard, if you read this, please, fix
your mailer with regard to threading or buy another one. Please.

[0] http://gcc.gnu.org/develop.html#stage3
[1] http://gcc.gnu.org/ml/gcc/2004-08/msg01418.html
[2] http://gcc.gnu.org/ml/gcc/2004-09/msg00803.html
[3] http://gcc.gnu.org/ml/gcc/2004-09/msg00857.html
[4] http://gcc.gnu.org/ml/gcc/2004-09/msg00853.html
[5] http://gcc.gnu.org/ml/gcc/2004-09/msg00453.html
[6] http://dberlin.org/gccwiki/
[7] http://gcc.gnu.org/ml/gcc/2004-09/msg00431.html
[8] http://gcc.gnu.org/ml/gcc-patches/2004-09/msg00333.html
[9] http://gcc.gnu.org/ml/gcc-patches/2004-09/msg00505.html
[10] http://www.gnat.info/
[11] http://gcc.gnu.org/ml/gcc/2004-09/msg00454.html
[12] http://gcc.gnu.org/ml/gcc/2004-09/msg00459.html
[13] http://gcc.gnu.org/ml/gcc/2004-09/msg00467.html


Mathieu
-- 
Mathieu Lacage <mathieu.lacage@sophia.inria.fr>