Re: Converting GCC to compilation with C++

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    But you manage to commit un-regression-tested code more often than not.

That's a bit of an exaggeration, as you know.  It does indeed happen, but
not often.

    Is your point that, if you, as a GCC maintainer with global write
    priv, manage to knowingly commit untested code, then MI should be left
    out?  That would be cynic.  That logic would imply that any useful    
    feature of C, as we currently practice, should be rejected.  Not just
    use of a C++ subset.

There's a big difference between *testing* and the issue of coding standards.
If something isn't tested and doesn't work, we'll pretty quickly realize that
it doesn't work and will get it either removed or fixed.

But there's no such analog with code standards issues, be it whitespace, code
comments, or choice of what features are used.  There's no testing process to
verify compliance and nothing that will "break" if the standards aren't
complied with.

So in some sense, I actually do agree with you: if even our rigorous testing
procedures mean that patches that don't work sometimes get checked in
(something that's hardly unique to me, of course), then how can we hope to
verify that each patch will meet coding standards?