At 19:17 Uhr +0200 12.04.1997, Hauke Fath wrote:
>Hi,
>
>a kernel built from -current sources of 09 April (ncrscsi driver with
>Allen's patches) never makes it to the copyright line (serial console) but
>hangs in the booter.
>
>A kernel with sbc driver built from 11 April sources stops dead in the
>booter with "Failed to set kernel symbol _boothowto" (the booter is Allen's
>v1.9.6).

Following up to my own mails - this one and the ncrscsi one:

An sbc kernel built after the recent SCSI changes exhibits the same strange
behaviour wrt. the Quantum LPS105S that I reported for the ncrscsi kernel
with Allen's patch.

The kernel was built on /var which is on the TRB850; after I copied it to
/stand which is in the root partition on the LPS105 I saw

root:~ # cmp -l /stand/netbsd.nfs.970413
/var/obj/sys/arch/mac68k/compile/WEIDE/
netbsd
 49153  71  40
 49154   0  71
 49155  14   0
 49156 137  14
 49157 210 137
 49158  40 210
 49159 153  40
 49160   0 153
 49161  30   0
 49162 110  30
 49163 156 110
 49164 376 156
 49165 354 376
 49166  40 354
 49167 160  40
 49168  14 160
 49169   0  14
 49170 116   0
 49171 220 116
 49172  44 220
 49173   0  44
 49174 130   0
 49175 117 130
 49176 146 117
 49177   0 146
 49178   6   0
 49179 362   6
 49180  57 362
 49181  56  57
 49182 376  56

[...]

That is, the same offset of one byte as reported for the ncrscsi kernel,
this time starting from 49153 which is close to the magic number 0xC000.

I dimly recall a report to current-users about selfmade kernels that don't
run; looks like a MI code bug. Shall I send-pr?

	hauke


--
"It's never straight up and down"     (DEVO)