Date: Tue, 26 Feb 2002 14:33:06 -0800 (PST) From: Matthew Dillon <dillon@apollo.backplane.com> To: Robert Watson <rwatson@FreeBSD.ORG> Cc: Garance A Drosihn <drosih@rpi.edu>, Alfred Perlstein <bright@mu.org>, Julian Elischer <julian@elischer.org>, Jake Burkholder <jake@locore.ca>, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: src/sys/i386/i386 exception.s genassym.c machdep.c mp_machdep.c mpapic.c swtch.s vm_machdep.c src/sys/i386/include cpufunc.h pcb.h src/sys/i386/isa apic_vector.s clock.c icu_vector.s intr_machdep.c intr_machdep.h npx.c src/sys/kern ... Message-ID: <200202262233.g1QMX6D98254@apollo.backplane.com> References: <Pine.NEB.3.96L.1020226160553.28921X-100000@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
:Perhaps I'm not recalling an earlier incident. The number of times where
:I've asked you not to commit something because something identical exists
:in a jhb tree is precisely 1. That's not a very big N. One hates to dig
:up long dead issues, but could you expound on what you are referring to?
Two that I would put in the aggregious category (UCRED, critical*),
one that was really quite annoying (the ucred NULLing code) and
ate half a week of both Julian and my time, the pool mutex
expansion argument which I would label as 'annoying', not to
mention a few others.
And then there is the constant spew of 'Matt doesn't know what
he is talking about' crap on IRC, which is really getting on my
nerves. I was doing SMP on 8 bit cpu's 18 years ago for gods
sake! And constant references to 'the plan' for this or
that, which is in John's head apparently, and totally undocumented
source code, or constant references to how 'other people do it'
-- a fine discussion point, but no excuse to not do something a
different way and certainly no justification to object to another
viewpoint. Hello! FreeBSD is NOT solaris.
Well, I've had it. I am not coddleing up to this crap any more.
If there were no interference I would be *done* with stage-2 by
now, instead of still waiting to get stage-1 in place. I would
have had UCRED completely instrumented by now, instead of still
waiting for a tiny patch to read-only functions in a single file
to get unstalled (or even for JHB to commit some of his stuff, in
which case I would like to be able to instrument Giant).
And I'll tell you something else, too. I'm not afraid to make
changes if I turn out to be wrong on something. I'm not afraid
to rip out and redo code. It's how I work best, in fact, and should
not be confused to 'working' vs 'non-working' code. Theory
works only up to a point, then you actually have to code it up and
have a lot of people play with it. Nobody who tries to engineer an
entire system all in one go will succeed, or certainly not succeed in
any worthwhile time frame. John is welcome to work that way but he
should not drag the rest of us (and specifically me) down with him.
Even the 4 hours that this commit was in the tree reaped benefits. I've
already had several people email me that it works, and Ian even found
a minor bug that can occur if a process switches tasks while in a
critical section. To me that is a huge reward, because I specifically
instrumented the code with a sysctl to turn it off so when Ian reported
that to me it took two seconds for him to verify that it was the
new critical code and another few seconds for me to locate and solve
the problem. Try that with a mega-commit and see how far you get!
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200202262233.g1QMX6D98254>