Date: Sun, 23 Dec 2001 12:33:12 -0500 (EST) From: Robert Watson <rwatson@FreeBSD.org> To: Greg Lehey <grog@FreeBSD.org> Cc: Bruce Evans <bde@FreeBSD.org>, Matt Dillon <dillon@FreeBSD.org>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org, FreeBSD Core Team <core@FreeBSD.org> Subject: Re: cvs commit: src/sys/dev/sio sio.c Message-ID: <Pine.NEB.3.96L.1011223120751.8511h-100000@fledge.watson.org> In-Reply-To: <20011223133156.I88202@monorchid.lemis.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 23 Dec 2001, Greg Lehey wrote: > (dons -core hat) > > Come on, people, this isn't appropriate behaviour. I'd have expected a > more mature approach from both of you. Please don't touch these sources > until we've had time to talk about it (which probably means at least 24 > hours). For whatever it's worth, here's my view: Bruce and Matt need to go sit in their respective corners for a few minutes, take a deep breath, and come back when they're ready. As I see it, there's plenty of fault to go around for everyone. Matt probably should have run the patch by Bruce before committing it: common courtesy dictates contacting the maintainer to give them a pass is a reasonable thing to do. A casual glance at the sio.c history reveals a fair number of Bruce's 'Reviewed by's, although they are not universal. As there was no rush in getting it committed (we've lived with the problem for ages, another 48 hours would have hurt no one), this seems a reasonable course of action. If Matt anticipated objections from Bruce, that suggests arch@FreeBSD.org as the place to take those concerns. Bruce should not have immediately backed out the fix -- an objection by e-mail would have been quite sufficient, asserting a maintainer bit and requesting a backout, and any technical concerns about the problems with the commit. A solution appears to be needed, whatever that solution may be, so taking the concerns seriously is important. I was actually quite surprised to see Bruce's immediate backout: Bruce has a long history of taking concerns with commits to the committers out-of-band, and often runs for years with local fixes to problems introduced on the main tree and never accepted back by the introducer of the problem. Likewise, Matt should not immediately have backed out the back out: he received a specific complaint from the maintainer of the code saying that solution was inappropriate. Feelings about the technical merits of the commit are not relevant in this situation, as it's clearly a conflict resolution scenario in a case where an administrative tool (MAINTAINER) is being used. Asserting the negligence of Bruce in fixing a problem doesn't help: he didn't present any evidence that he'd previously raised the problem, making a rapid and unreviewed commit of a fix unjustified. Even assuming Bruce's commit was agressive and un-called for (reasonable conclusions), that response takes Matt down to Bruce's level, in my eyes. It was clearly retaliatory, and inappropriate. It seems like a reasonable conclusion, therefore, that Bruce and Matt both acted inappropriately. I don't really have any opinion concerning the technical fix: there are almost always multiple valid ways to solve a problem Bruce's opinion certainly counts for something, as the maintainer of the code and as one with lots of experience using and modifying the code, and Matt's opinion counts due to specific material problems experienced. There was no need to rush the commit, the backout, and certainly no need to back out the backout. I think we all believe there's a problem with what is there right now: on at least two -CURRENT i386 boxes I have, I frequently experience silo overflows. I haven't experienced them in -STABLE lately, but certainly used to. And Doug Rabson has reported similar problems on the Alpha platform. Even if the real fix is to improve interrupt latency problems, we will need a temporary fix, as some of those interrupt problems appear to be inherrent to the gradual SMPng development process we've adopted. In the future, I'd appreciate it if both Matt and Bruce would wait at least 12 hours before backing out changes under circumstances like these. Preferably 48 hours. And take a deep breath before getting involved in this sort of argument. Under such circumstances, it's rare for any one person to be 100% right, which means that acting righteously to liberate source code from the hands of an oppressor (either one) is inappropriate at best, and really quite disruptive in practice. Robert N M Watson FreeBSD Core Team, TrustedBSD Project robert@fledge.watson.org NAI Labs, Safeport Network Services 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?Pine.NEB.3.96L.1011223120751.8511h-100000>