phil@unicorn.wwu.edu (Phil Nelson) (10/11/90)
I have used the changes that appeared in the pc532-src mail list to
get a new version of minix. I have used that version for a while and
thought I should post my experience.
a) The patches went great, no problems.
b) The RTS/CTS for the rs232 lines was good.
c) I still had problems with the rs232 lines. No characters were lost
and the system did not crash, but it lost intterupts (I think that
is what it did) and required more key strokes to cause previous ones
to be recognized and sent. It was very pronounced when using kermit.
e) The system was generally more unstable than Bruce's kernel. I had to
reboot the system due to hangs etc about a factor of 10 times more.
f) The bpt trap did not return me to the monitor. I don't know what it
did.
g) Also, if kermit (or other software) opened a port that did not have
CTS asserted, you could not kill the process until CTS was asserted.
In conclusion, some of the changes were nice, but I have switched back to
the original distribution due to the high number of system hangs.
--Phildlr@daver.bungi.com (Dave Rand) (10/11/90)
[In the message entitled "minix532 changes" on Oct 10, 13:51, Phil Nelson writes:] > In conclusion, some of the changes were nice, but I have switched back to > the original distribution due to the high number of system hangs. > Funny - George and I switched to the new code due to the high number of system hangs with the old code :-) Anyone else had problems? We are using our system with two ports active with users, one printer attached, and another port coming in from wombat for uucp work. Seems to work well, and runs for weeks with no hangs... -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
sverre@lars.Seri.GOV (Sverre Froyen) (10/11/90)
> >Funny - George and I switched to the new code due to the high number >of system hangs with the old code :-) I agree, the tty driver is less apt to hang after applying the changes. Especially when using two ports on the same DUART. > >Anyone else had problems? We are using our system with two ports active >with users, one printer attached, and another port coming in from wombat >for uucp work. Seems to work well, and runs for weeks with no hangs... > I have had uucp hang the port (and system) several times when doing large file transfers from my ICM3216. I also have an occational disk problem. This happens with both the old and new kernel and goes roughly as follows: SCSI error reporting drive not ready (usually just one). SCSI error reporting bad command (block) (usually several). SCSI timeouts (several). Somtimes the problem clears itself and somtimes I need to power-cycle to clear it (hitting the reset button is not sufficient). If anyone else have seen something like this I should like to know (to rule out a bad disk). Sverre -- Sverre Froyen sverre@seri.gov, sunpeaks!seri!sverre
dlr@daver.bungi.com (Dave Rand) (10/13/90)
[In the message entitled "minix532 changes" on Oct 12, 16:41, Phil Nelson writes:] > Also, bpt > would not get me back to the monitor. Did anyone have that work for them? > Well, <ahem>, you sorta-kinda-hafta make this small, little, tiny change to the monitor... Right before it goes to print a character in the tty driver in the monitor, you have to write a 0x4 to the command register to re-enable the transmitter... BPT works, but when the monitor trys to print out the message, it can't because TXRDY isn't set, because the transmitter is off. That should fix up you BPT issue... I'll look into the other stuff this weekend (I didn't use kermit much at all...) -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
phil@unicorn.cc.wwu.edu (Phil Nelson) (10/13/90)
>>Funny - George and I switched to the new code due to the high number >>of system hangs with the old code :-) > >I agree, the tty driver is less apt to hang after applying the changes. >Especially when using two ports on the same DUART. I also agree about using two ports on the same DUART. The big problem I had was having a terminal on one port and trying to kermit out another port to a unix system. No characters were lost, but often a character was not send until another chacter was hit. That made it very unusable to talk to a screen oriented editor and many other things. Also, bpt would not get me back to the monitor. Did anyone have that work for them? --Phil