[comp.sys.amiga.tech] DNET tty wierdness

bmacintyre@watdragon.waterloo.edu (Blair MacIntyre) (11/30/89)

When I run DNET, the tty's that it creates for fterm aren't set up
properly:

watdragon [101]% ls -l /dev/ttyq9
crw-rw-rw-  1 root           20,  25 Nov 29 13:05 /dev/ttyq9

There should not be rw perms for everyone, first.

Second, why is the owner root?!?!

Any ideas?
-- 
-- Blair MacIntyre, Professional Leech on Society ( aka CS Graduate Student )
-- bmacintyre@{watcgl, watdragon, violet}.{waterloo.edu, UWaterloo.ca}
-- Dating, verb: prearranged socializing with intent.

hull@hao.ucar.edu (Howard Hull) (12/04/89)

In article <18796@watdragon.waterloo.edu> bmacintyre@watdragon.waterloo.edu (Blair MacIntyre) writes:
>When I run DNET, the tty's that it creates for fterm aren't set up
>properly:
Or at least for your purposes!
>watdragon [101]% ls -l /dev/ttyq9
>crw-rw-rw-  1 root           20,  25 Nov 29 13:05 /dev/ttyq9
>There should not be rw perms for everyone, first.
Yup.
>Second, why is the owner root?!?!
>Any ideas?
Oh yeah, ideas, I got.  Actually when the program attempts to open an
FTERM 8195 (unnamed) server, it tries to set the permissions of the port.
But since the port is owned by root, it will fail unless the user who opened
the device is also root (or has obtained access via a process that is suid
root).  This doesn't much bother Matt Dillon, who, as you might surmise,
has access to root and suid root processes on his machine (you gotta bet!).

For some Unix systems an error will be returned to the user when the chown
fails; of course, if the chown fails, the chmod will never be called.   The
Sun 4 system I work on cheerfully tells me that the chown failed, and then
goes on as if nothing much important was happened.  This has had some hilarious
results from time to time, as often the Unix system will subsequently allocate
the port to a Sun workstation user, and then interleave our byte streams.
Although this will not do much to the DNet RDPPs, it will totally hose the
Sun workstation, and will shortly thereafter hang the machine's entire terminal
I/O system.

So if you find yourself in this situation every day you come to work, you
first gravitate to asking for a named port (invoke dnet as "Run DNet -s" plus
whatever other options you want, and then after starting dnet on the Unix end,
use the CLI to ask for a named port via the command "Run FTerm 8193".  After
a while you get tired of even that, and if you're not too red hot at setting
up compilation environments aligned to emulating Matt's (i.e., use sup32lib
and the rest of his environmental constituents), you can actually quite easily
patch your FTerm executable with NewZap to replace the 8195 in there with 8193
(so Matt, don't put another 8195 constant in there anywhere or I'll get royally
confused!).  Then you can routinely run the thing without the -s flag and the
need to call FTerm from the CLI.  Not only that, but it will open the FTerm
with the more appropriate permissions you are looking for, people will be able
to send you mail and reminders, AND you'll be able to subsequently set the
permissions however it may please you (i.e., using mesg and biff).  Of course
then you won't be able to read news during the day without being spotted by
the Unix w command like you could if you were able to chown the port supplied
by the 8195 server, but what the heck, who did that anyway???  :-)

>-- Blair MacIntyre, Professional Leech on Society ( aka CS Graduate Student )
>-- bmacintyre@{watcgl, watdragon, violet}.{waterloo.edu, UWaterloo.ca}
>-- Dating, verb: prearranged socializing with intent.
						Howard Hull
						hull@ncar.ucar.edu

keithh@atreus.uucp (Keith Hanlan) (12/06/89)

In article <18796@watdragon.waterloo.edu> bmacintyre@watdragon.waterloo.edu (Blair MacIntyre) writes:
>When I run DNET, the tty's that it creates for fterm aren't set up
>properly:
>
>watdragon [101]% ls -l /dev/ttyq9
>crw-rw-rw-  1 root           20,  25 Nov 29 13:05 /dev/ttyq9
>
>There should not be rw perms for everyone, first.
	I don't know much about Dnet but these things tend to be vendor
	specific. On my Sun, consoles a rw- owner and rw- others and group
	depending on umask and mesg. On the HP, it's similar but they don't
	even bother with read permission for group and other. Check your
	umask and your mesg.
>
>Second, why is the owner root?!?!
	On my Sun, consoles are owned by root. On my HP, consoles
	are owned by me (keithh).

	I think its just that Matt is trying to avoid changing the
	conventions of the particular OS, in an attempt to be as portable as
	possible.

	Regards,
	Keith
Keith Hanlan
Bell-Northern Research, Ottawa, Canada 613-765-4645
uunet!utgpu!bnr-vpa!bnr-fos!bmers58!atreus!keithh or keithh@bnr.ca

rjtatz@hpuxa.ircc.ohio-state.edu (Robert J. Tatz) (12/06/89)

Two questions about DNET:
1) Has anyone got it running on a HP9000 running unix?
2) Could an A2500 with the CBM 7 port serial board 
   be used as a file server for 7 A1000's running DNET: ?
-Bob         rjtatz@hpuxa.ircc.ohio-state.edu