Re: [Gdbheads] proposed change to GDB maintainership rules

[Ian Lance Taylor]

Jim Blandy <address@hidden> writes:

> Ian Lance Taylor <address@hidden> writes:
> > I've already been involved in one situation where the maintainers
> > could not come to a consensus, and it lead to the gcc/egcs split.
> > That is probably the worst possible solution, although it turned out
> > OK after a couple of years.  Making it turn out OK involved an
> > enormous amount of time on the part of several individuals who worked
> > steadily to try to resolve the issue.  I recommend against this
> > approach.
> 
> When you say "I recommend against this approach", does that mean that
> you feel the GCC community had a better alternative at the time the
> split was made?  That, in retrospect, they would not have started
> EGCS?  I'd like to hear more about how you view the choices in
> hindsight.

That's a good question, and, no, I don't think we had a better
alternative.

At the time, as you know, Richard Kenner was the sole gcc maintainer.
The job was clearly too big for him, and he was increasingly
unresponsive and difficult to work with.  He also worked for GNAT, as
indeed he still does, and he prioritized patches which helped GNAT
even when they hurt everybody else.  RMS strongly supported Richard
Kenner, I believe in part because all the other significant gcc
contributors worked for Cygnus, and RMS was concerned about Cygnus
domination of gcc.  The situation was untenable for at least a year
before we finally broke away in 1997.

Obviously, I'm biased, since I was firmly on one side of the split.

I think the key elements which led to the split were:

1) gcc had only one maintainer, Richard Kenner.
2) Kenner was unresponsive--he blocked patches without explaining how
   they might be rewritten.
3) Kenner was not an independent voice--he followed the interests of
   his employer even when they went against the interests of others.
4) As gcc maintainer, Kenner answered only to RMS.
5) RMS refused to add another gcc maintainer, since that person would
   have had to come from Cygnus--there were no plausible alternatives.
   (RMS might have accepted me as a maintainer, since I had a proven
   track record of GNU maintenance, but I had to turn that down since
   I did not know gcc well enough).
6) There was no other court of appeal.
7) Cygnus had a strong commercial interest in gcc, and was thus
   strongly motivated to solve the problem somehow--that is, Cygnus
   could not just walk away, which is what most other gcc contributors
   had already done.

We started the split with the clear intention of becoming the mainline
of gcc development, and replacing the existing mainline, and, indeed,
that is exactly what happened.

Ian