Re: [patch rfa:doco rfc:NEWS] mi1 -> mi2; rm mi0

[Andrew Cagney <ac131313 at redhat dot com>]

> This is only partly true.  We are actually pretty conservative about changing command output.  We haven't broken annotate for a while, IIRC.  The mi is more like the command output, and I think we should have the same level of conservatism about this.
One of the motivations behind the MI output was that a parser could be 
written in a way that allowed it to discard anything it didn't recognize.

For instance: additional events, or additional output fields should both 
simply be discarded.  They shouldn't be viewed as MI changes.

The thing that is triggering mi2 is the changes to actual responses 
(breakpoints as you pointed out but also others.).

Anyway, at the time of each GDB release, a decision should be made about 
freezing the current MI interface and starting a new one.  Here, we've 
frozen mi1 and moved onto mi2's development.  At some point in the 
future mi2 will be frozen and development will move to mi3 (I strongly 
suspect it will be pretty quick - 5.4/6.0).  After a freeze, the old 
syntax should hang around for a limited period of time.

The pratical translation of this is  ``how long before the mi1-*.exp 
files can be deleted''?  My guess is two releases - here 5.3 and 
5.4/6.0. That means that 5.5/6.0 (in roughly a year) there will be 
released a GDB  that doesn't support "mi1" -- for GDB I think that is a 
long time!

Andrew