This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Implementing multi-methods for C++ in GCC
I strongly suggest that you have at least a draft standard before you
start, and that you think seriously about corner cases and interaction
with other language features before you start coding (and let others
try
to find flaws with it as well).
Thanks for your comments and encouragement to pursue this path. It
seems we have the option to actually run this past the standards
committee members earlier rather than later, so we will make sure we do
so. We are familiar enough with the standard that it shouldn't be a
problem.
I continue to hope the implementation is 95% run-time library matter
and relatively little needs to be done in the compiler proper, but that
doesn't make it any easier to specify in language terms :-) At this
point I can only say that I have thought through the problem enough to
have realised how much I haven't yet considered.
The fact that this hasn't always been
done has contributed to lots of problems with GCC extensions in the
past,
and the implementation will be cleaner if you identify as many issues
as possible up front.
Agreed.
Lassi
--
If that makes any sense to you, you have a big problem.
--C. Durance, Computer Science 234