On Aug 2,  3:22pm, Markus Baeurle wrote:
> Subject: Re: Compiling a new kernel?!
> On Thu, 1 Aug 1996, Ale Terlevich wrote:
>
> > > the core team of NetBSD). I hope that Mark will find some time to
> > > check the latest changes in soon.
> >
> >   I was under the impression that only bugfixes would be checked in until
> > NetBSD 1.2 is finally released, then NetBSD-current will continue as
> > before with all of the latest goodies :)
>
> It's NetBSD-*current*, isn't it?
> Anyway, may I quote Mark to clarify this:
>
> MB> Date: Fri, 19 Jul 1996 12:45:05 +0100
> MB>
> MB> The source for the 4444 kernel will be checked into NetBSD-current
> MB> shortly.
> MB> As for the 1.2 release kernel, technically if should be the kernel source
> MB> frozen with the NetBSD 1.2 source. This will be the base kernel for
> MB> the 1.2 release.
>
> >   Wasn't there some problem with X under these new kernels anyway?
>
> Well, I use bsd-4444 together with Xarm18a and don't really have that much
> problems. In fact, it often works flawlessly.
> You get some error messages at startup, but that's it.

Yes, there is a problem (a major one) with running Xarm with versions less than
20 under new kernels which can cause both the server and applications to
randomly core dump for no apparent reason. This is to do with an interrupt race
or two and is largely dependant on the interrupt timings (ie the devices you
have connected) so some people will rarely see and others will see it a lot.
Things are likely to get steadily worse for Xarm < 20 and corresponding
libraries as kernel interrupt handling improves.

> I didn't check recently if the menus of Athena-based programs always don't
> work, though. I remember that I had problems with that, but I can't remember
> offhand when they first showed up. There has always been the problem that
> didn't work occasionally, and with always I mean from the time RiscBSD
> became first available.

There are three problems with menus which people tend to confuse.

a) If you switch virtual terminals X does not see the key-up event for the alt
or function key you used to switch out. If you switch in with the other alt key
then X thinks that the original alt + function key are still pressed (it sees
an alt release event, but for the other alt). The function key 'getting stuck
down' usually doesn't matter much but the alt key is a modifier so it will
change the affect of mouse clicks. The work around is that if you see the mouse
behaving oddly press and release both alt keys just in case they are flagged as
pressed.

b) The original X servers had auto-repeat on the mouse buttons under some
circumstances (oops). This caused intermitent problems with menus if you
weren't quick. It disappeared by about Christmas (overlap with problem c).

c) The gcc 2.7.2 released with the beta release had a bug in it which caused it
to miscompile things. The X set distributed in November was compiled with it
and it messed up menus in anything built with the athena widget set (including
just about all window managers). This was fixed in February but due to a screw
up the sets didn't make it to the ftp site until April and then only as an
upgrade set. I just checked and it appears that the base X sets in the beta-1.1
dir are still the broken ones from November (datestamped Feb). Any applications
compiled without the upgrade should be regarded as suspect. This includes any
sets on the ftp site with any files dated before 4th April (eg fvwm, ctwm, tcl,
tk, gs262, gview,... - yikes, I didn't realise so much hadn't been rebuilt).

I have given Mark the X base distribution for the 1.2 beta. This is built from
the X11R6.1 source tree and has quite a few fixes in it. He will turn it into
sets when he has time. One of these days when I manage to wake up *before*
leaving for work in the morning I hope I will remember the tape with fvwm, xpm
and bits of contrib on it.

Cheers

Rob

--