more on the C++ abstraction penalty

[Mark Mitchell]

>>>>> "Joe" == Joe Buck <jbuck@racerx.synopsys.com> writes:

    Joe> Furthermore, you haven't announced any kind of schedules yet,
    Joe> eg when is the feature freeze?  I'm still seeing much more
    Joe> ambitious changes wrt the preprocessor being discussed
    Joe> without being similarly shot down.

That's a fair criticism.  I don't want to see any ambitious changes
anywhere.

It's also fair to note that I haven't announced schedules.  Part of my
problem is that I have found it very hard to obtain scheduling
information out of various people about completing features that
seemed essential.

That said, I have been encouraged privately to set a schedule so as to
try to motivate people, and I intend to do that soon.  We are changing
to V3 at the end of this week, and I hope to change to the new ABI
within a week or two of that.  At that point, I intend to go to slush.
I have hoping that GCC 3.0 will be ready in Q1 2000, and I think that
is doable.  It will mostly be a feature release; I expect GCC 3.1 to
contain many fewer new user-visible features, and many more
optimizations (such as the one you've been discussing).

N.B.: the Java front-end is still not compatible with the new ABI.  I
have been encouraged by some people (including people at Red Hat) to
ask the SC to drop Java from the GCC 3.0 release.  It is frustruating
to me that I brought this issue up with the Java team at least six
months ago, and that there has been (to my knowledge) little progress.
My current thinking is to set a schedule for 3.0, branch, and let the
Java people decide whether or not they want to update gcj on the
branch or not.

Your flammable R.M. :-),

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com