Re: Obsolete XXXX

[Andrew Cagney <ac131313 at cygnus dot com>]

> I understand the motivation of trying to get things converted to
> multi-arch, but things should be declared obsolete because the
> chip is dead or because there are no users, not because there are
> no volunteers to change the code in a way you'd like it to be
> changed.  The usual practice in GCC has been for someone to take
> a deep breath and then do a mass change of the sources; if you
> look at the GCC Changelog for the past year, you'll see some
> major updates that not only involved all the obscure targets, but
> were tested on them as well.


The process being followed to get targets obsoleted is:

o 
announce the candidate on gdb@
	(to see if anyone steps forward)

	wait

o 
announce the intent on gdb-announce@

	wait

o 
mark the code as obsolete

	wait

o 
make a GDB release

o 
remove the obsolete code

I'm only at the first step!

BTW, for changes across targets, GDB's strategy is similar.  The 
developer should at least ensure that all targets continue to build and 
start.

FWIW, the problem with doing this for multi-arch is that the process, 
unless tested, is pretty much guarenteed to break the target.  My memory 
of the previous disussion on this drew the conclusion that obsoleting a 
target would be better than converting it blindly.  This way we (GDB) 
wouldn't end up wasting our time trying to update a broken target.

enjoy,
Andrew