GCC RM Q&A, C++ in GCC

[Mark Mitchell <mark at codesourcery dot com>]

Thanks to all who participated in the GCC Release Manager Q&A on IRC
earlier today.  Joseph Myers logged the chat session and will put it on
the GCC Wiki soon.  There were a lot of good comments and questions.

Thanks especially to Richard Guenther and Joseph Myers who both
accommodated my unilateral setting of the schedule for the chat.
Unfortunately, Jakub had a conflict and could not attend.  The consensus
seemed to be that this was a useful discussion, so we will do it again
in a month or two; I will announce a date soon.

The issue of using C++ to build GCC itself came up again in the
discussion.  Although there is certainly not unanimity, there seems to
be an overall consensus in the GCC developer community that limited use
of C++ would be beneficial.  That consensus was present today on IRC,
present in previous email discussions on this list, and present at the
GCC Summits over the past two years.

Therefore, I've asked the SC for permission to do require a C++98 host
compiler to build GCC.  Once that permission has been obtained, we can
then work on what coding standards are appropriate.  My expectation is
that we will limit use of C++ features at first quite strictly; for
example, we might forbid use of virtual functions, or multiple
inheritance, or overloading.  But, we'll take advantage of some obvious
benefits, like use of std::vector in lieu of VEC.h.

I will report further once the SC discussion has reached a conclusion.

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713