On 4 Jun 2001, John Hasler wrote: > The LSR test will not be performed if ASYNC_BUGGY_UART is set, and > include/linux/serial.h says: > > #define ASYNC_BUGGY_UART 0x4000 /* This is a buggy UART, skip some safety > * checks. Note: can be dangerous! */ > > Try configuring the port with 'skip_test'.
It has worked with an explicit 'skip_test' and the implicit test, with both the 2.2.15 and 2.4.3 kernels. I think it is going to take something more rigorous than the simple 'change one variable at a time' approach I've (mostly) been using. The last time I noticed really strange behaviour like this[1] was with KDE (the "restart KDE" or "rm ~/.kde" days), and it appeared that inconsistent configs was contributing the problem (hard-coded, system-wide, per-user, and DB could all be different, depending on what you had been upto), simply because there was always wider agreement between the pieces as to the way things should be when they worked than when they didn't work. All this user friendly config handling is nice... but it appears to be pretty fragile at times. - Bruce [1] I reboot whenever unstable gives me a new libc or xlibs (to ensure upgrade related problems bite me sooner rather than later), and tend not to fiddle with it if it is not broken. The only thing I can think of that may have had a delayed effect was if something like setserial came along and fiddled (or let me fiddle) with things I would only see the actual effect of when I rebooted (like maybe changing from "manual" to "autosave-once" when that nice S-lang debconf screen comes up 'cause setserial got upgraded).