>> If you're (reasonably) sure that it's NBSD related (vs. something in
>> the boot ROMs!), [...]
>
>> I just anted to know if this was a "update your boot ROM" issue or a
>> "NetBSD bug-fix"...
>
>Well, let me be precise.
>
>I've seen behaviour that convinces me there is a bug in the kernel in
>use on my bwtwo-console machine.  (Specifically, in the RASTERCONSOLE
>code for shifting part of a line left.)


I have verified the behaviour on two IPC boxes (bwtwo with 
RASTERCONSOLE).  Both are running 1.6 ROMs though AFAIK, this
should *not* affect the problem (?)

Note, however, that I have also noticed the behaviour when
*SHIFTING RIGHT* as well as left!

So, for example, type "asdfghjkl" on the command line.
With the cursor at the end of the line, deleting *left*
works as expected -- the rightmost character disaappears and
whitespace takes it's place.  The cursor moves left one
character position.

However, if you use ARROW_LEFT to move cursor to the middle
of the text and then try to delete, the character is deleted
*but* the characters to the right of the cursor get "smeared"
as they are shifted to the left to reflect the missing character.

Note that it appears that you have to have a few characters 
to the right of the repositioned cursor for this to be noticed
(I haven't done any exhaustive testing since keeping the
monitor attached to any of these machines is problematic -- no
place to locate it... hence the reason I run headless!)

Now, if you also position the cursor in the middle of the
text and try *adding* text, the characters get smeared
as they are shifted *right*!

Note that in any of these cases, ^R reveals that the
correct characters are present on the line -- just the
glyphs are distorted!

>The description you've given so far is consistent with the behaviour
>I've seen.
>
>Now, whether there is also a similar bug in some ROM versions, I don't
>know; I've seen no behaviour that makes me think so.  Whether the
>behaviour you're seeing is due to the same bug as the behaviour I'm
>seeing, I also don't know, though I suspect so; I'm treating this as a
>working hypothesis.


From the OFW prompt, there doesn't (?) appear to be a way to
move the cursor to the left *in* the command line (i.e. ARROW_LEFT
isn't recognized as such).  I haven't built a kernel without
the RASTERCONSOLE option to determine if the behaviour would
also be present with the ROM'ed blitter.  If it is ROM version
dependant, I figure I could only offer that *one* (v 1.6)
datapoint, anyway...

I suspect that the previously mentioned BLINK problem may
also be ROM related.  I have verified that the two Classic's
that DON'T blink are running 2.9 ROMs while the one that
*does* blink runs a 2.12 ROM.

First order of business is to track down the "latest/greatest"
version of each firwmare release and burn a new set of ROMs
to ensure all machines are at the same release level.

Of course, it is also possible that the PCB revision levels may
differ.  :-/

Thx,
--don