Date: Sun, 28 Oct 2001 11:13:31 -0500 From: Sergey Babkin <babkin@bellatlantic.net> To: Poul-Henning Kamp <phk@critter.freebsd.dk> Cc: Greg Lehey <grog@FreeBSD.org>, mjacob@feral.com, Jonathan Lemon <jlemon@flugsvamp.com>, "Andrey A. Chernov" <ache@nagual.pp.ru>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: Causing known breakage (was: cvs commit: src/sys/kern kern_conf.c subr_disk.c) Message-ID: <3BDC2EAB.A59C7071@bellatlantic.net> References: <33623.1004269277@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
Poul-Henning Kamp wrote: > > In message <20011028205917.C88146@monorchid.lemis.com>, Greg Lehey writes: > > >> We never had such a rule, and infrastructre changes would be close > >> to impossible to perform if we did. > > > >That's a claim you continue to make. It makes life (or at least your > >arguments) easier for you, but I disagree, and I'm not the only one. > > "We are many who think that..." I am another one of these many. I believe that almost any kind of externally visible infrastructure change must be done in a backwards-compatible way. That is, with providing translation from the old infrastructure to the new one. > >> Proof: Look at these messages from LINT: > >> WARNING: COMPAT_SVR4 is broken and usage is, until fixed, not recommended > >> #warning "The eni driver is broken and is not compiled with LINT" > >> #warning "The fore pci driver is broken and is not compiled with LINT" > >> #warning "The lmc driver is broken and is not compiled with LINT" > >> > >> This basically says that SCO/SVID compatibility, one entire ATM stack > >> and a T1/E1 driver have been shot. > > > >Who did it? > > Can't remember, seem to recall it was related to newbus/PCI/interrupt > or something in that area. Someone has decided to remove the (existing!) support for the legacy PCI infrastructure. > If you want to implement this new rule of yours, I suggest you do Mike Smith and I had a discussion in -hackers about a close issue and Mike has stated too that we already have this rule. > so chronologically rather than to pick on me. You active at breaking this rule, so Greg's picking is justified. > >Have you evidence that it wasn't done because somebody made changes > >without taking these issues into account? Currently this looks like > >an argument for my point of view. If we carry on like this, > >everything except the core functionality will be broken. > > Actually I think it is a neat thing. If nobody fixes these drivers And what exactly is neat about it ? > before 5.0-R we can obviously remove them then because they are > clearly NWOV[*] material in that case. > [*] "Not Wanted On Voyage", mark put on luggage which could be put > in the hold on trans-oceanic ships. That's a wrong assumption. This means only that nobody has enough interest (yet?) to start figuring out what change in the other parts of the system has caused the breakage and rewriting the drivers around it (and then re-testing them and fixing the new bugs). People who use them may prefer to stay with 4.x or say "screw FreeBSD" and switch to Linux. -SB 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?3BDC2EAB.A59C7071>