Subject: 1.5R and Xhpux
To: None <port-hp300@netbsd.org>
From: Jarkko Teppo <jate@UWasa.Fi>
List: port-hp300
Date: 02/05/2001 12:48:26
Hello all!

After several days of compiling my HP is finally at 1.5R.
A couple of observations are in order:

Softdeps can panic the system pretty efficiently. I was using a
1.5Q kernel (from the snapshot) and compiling a 1.5R generic and
when stripping the kernel it paniced. Softdeps were in the trace but
I didn't write it down, sorry:) It *might* be related that the 
default shell limits are not adequate for stripping.. don't know
(now there's a sentence)

Also I normally don't want to use XhpBSD and with the ELF transition
I'd have to use aout-emulation so I installed Xhpux and proceeded
with panicing the system, again. It's seemingly random *but* it 
only happens at server exit, it *sometimes* happens on normal
exit and *always* when you go to DDB from X. It also seems to 
die randomly.

panic: AMAP_B2SLOT: invalid byte count
Stopped in pid 2993 (Xhpux) at cpu_Debugger+0x6:   unlk a6
db> t
cpu_Debugger()
panic()
amap_splitref()
uvm_map_clip_start()
uvm_unmap_remove()
uvm_unmap()
grfunmap()
hpuxgrfioctl()
spec_ioctl()
vn_ioctl()
hpux_sys_ioctl()
syscall()
trap0()

Otherwise I really like -current. Sushi is nice (looks all wrong in
TERM=hp300h though), locale-stuff is good to have and UBC speeds up
the compiles even on a memory starved machine like mine.

Oh, the machine is a 380 with 28MB mem + random SCSI disks.
-- 
jht