Date: Fri, 12 Sep 2003 13:20:49 +1000 (EST) From: Bruce Evans <bde@zeta.org.au> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: cvs-all@FreeBSD.org Subject: Re: cvs commit: src/etc/etc.sparc64 ttys Message-ID: <20030912124508.T1009@gamplex.bde.org> In-Reply-To: <2113.1063314310@critter.freebsd.dk> References: <2113.1063314310@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 11 Sep 2003, Poul-Henning Kamp wrote: > In message <20030912063325.A1562@gamplex.bde.org>, Bruce Evans writes: > > >Other uart(4) bugs in this area: > >- the cua* devices are named uart*. I think ps(1), etc. strip off the > > first 3 characters of device names, so uart0 becomes t0. > > Yes, this is an ancient and bogus convention which will wreck havoc > in a number of surprising ports of broken. This is a necessary convention since ps and other output is in columns and has a limited width. > >The initial state devices or equivalent functionalty are more important > >than the cua devices IMO. > > If it is possible to set cua behaviour using stty (on the initial > state device) cua devices would be unnecessary and should probably > be avoided. Setting cua behaviour for non-bidirectional use is possible but not supported. Setting clocal is a fair approximation (cua behaviour is to ignore carrier on open (not just first open) but respect it later so that carrier drop hangs up unless clocal is set). Full bidirectional cua behaviour is not possible without actual, separate cua devices. It saves having a daemon to manage serial device accesses if it is actually required. I've never actually needed it. > >Appart from this, all devices on device driver foo should be named > >foo<number>, and no driver names should be longer than "foo". > > While 17576 device drivers would be enough for some time, I am not > sure we have the necessary mechanics to control or enforce them > being unique and I resist having to come up with a bogus TLA name > for a device in the cases where the vendor has already established > a perfectly good bogus name of a diffrent length for their device. > > See for instance sys/dev/musycc I resist using such vendors :-). Another annoyance is long driver names causing loss if info in systat interrupt output. "atkbd0 irqN[N]" didn't fit and caused the loss of "irq" for all devices. "atapci2 11" still doesn't fit and is truncated to "atapci2 1". Long names like "atapci" are easy to get if the driver builds names using simplistic concatenation of components (it is <driver><bus>). "atkbd" is essentially the same mistake (it is <pseudo-bus><driver>). Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030912124508.T1009>