[9fans] segattach plan9port

[9fans] segattach plan9port

[fernanbolando]

last night I was looking into vt source. I tried to compile under
p9p and in consctl.c it was calling the p9 system call segattach.

I was wondering if somebody can point me in a p9p port that shows
a very simple and straight forward example of the conversion of the p9 
to p9p segattach.

I did a quick google and I am currently browsing the source differences 
between p9 and p9p, but so far no luck yet.

just a note
Page in the browse source section of p9p site still points to the psv 
script, instead of the p9 port of Page.

Re: [9fans] segattach plan9port

[erik quanstrom]

shmemget(2) would be a more-or-less direct replacement.

really converting it would require using the thread library and send and recv
or another mechanism.  using a shared memory segment for the console state
doesn't feel right to me.

- erik

Re: [9fans] segattach plan9port

[Russ Cox]

> last night I was looking into vt source. I tried to compile under
> p9p and in consctl.c it was calling the p9 system call segattach.
> 
> I was wondering if somebody can point me in a p9p port that shows
> a very simple and straight forward example of the conversion of the p9 
> to p9p segattach.

As others noted, you don't really want segattach.
That code could be rewritten to make consctl a 
separate shared-memory thread instead of allocating
this one-time shared data structure.

But even if you do that, consctl depends on getting
terminal change (rawon, rawoff) events on /dev/consctl,
which doesn't exist on Unix.  To really port vt you'd have
to add all sorts of pty-handling code like in 9term.

If you need a cursor-addressed terminal window, 
you're better off just using xterm.  The cursor-addressed
programs you are trying to run will probably work 
better anyway.

Russ

Re: [9fans] segattach plan9port

[Derek Fawcus]

On Tue, Nov 27, 2007 at 11:03:02AM -0500, Russ Cox wrote:
> 
> But even if you do that, consctl depends on getting
> terminal change (rawon, rawoff) events on /dev/consctl,
> which doesn't exist on Unix.

I've been working on and off on a 'pty slave' which sort
of does that giving cooked/raw/cbreak messages based on
pty changes.  The idea being to convert the pty handling
crud to a message protocol.

The code is ugly,  being a wip which has grown as I
experimented, but does seem to work on Solaris/Linux/OSX.

The major missing bit is support for completion,  and then
the proper driver program.  One thought was to drive it
from a suitable hacked p9p 9term.

DF

Re: [9fans] segattach plan9port

[Fernan Bolando]

On 11/28/07, Russ Cox <rsc@swtch.com> wrote:
> > last night I was looking into vt source. I tried to compile under
> > p9p and in consctl.c it was calling the p9 system call segattach.
> >
> > I was wondering if somebody can point me in a p9p port that shows
> > a very simple and straight forward example of the conversion of the p9
> > to p9p segattach.
>
> If you need a cursor-addressed terminal window,
> you're better off just using xterm.  The cursor-addressed
> programs you are trying to run will probably work
> better anyway.

Actually the reason I was looking into the vt source was I needed a
terminal only
access to the 9grid. After a bit of patience you guys came up with
comterm, which is what I wanted.

Right now that I am still using 9pclient and
drawterm to to do some file transfer stuff, But later I will try to compile
comterm and switch to it.





-- 
http://www.fernski.com