[9fans] drawterm

[9fans] drawterm

[Russ Cox]

I am almost ready to declare dt2k "stable".

The random lock-ups were due to suspect X11 use
that I cleaned up a few weeks ago.

The fcp bug I mentioned before turns out to be a
problem with Plan 9, not with drawterm.  The bug was
in devssl (or in devmnt, depending on how you look at it),
not in drawterm.  There is a new /sys/src/9/port/devssl.c
and devmnt.c on sources.

There are new drawterm binaries at http://swtch.com/drawterm/

If you build yourself a kernel with the new devssl/devmnt
and are running the latest drawterm and see it crash or
do something else wrong, please let me know.

The latest drawterm has one new feature: if you are running
a plan9port factotum, then drawterm will use it.

Russ

Re: [9fans] drawterm

["Nils O. Selåsdal"]

Gabriel Diaz wrote:
> 
> Hi
> 
> I just downloaded the windows version and seems that this
> new version can't be used with the same params of the old one.
> 
> I try to run it without params and with the common /? -h --help,
> and i only get a crash. I think it try to open some environment
> variables that are missing. (i can send you the complete
> dump from windows if needed)
http://asgaard.homelinux.org/plan9/dt2k-mingw.exe doesn't crash atleast
if given /?, -h or --help

> How can be used this new version? are there some env vars I must
> use?

run it with the flags
-a the.auth.server -c the.cpu.server
(-u user and -s secstore can be specified as well)

Re: [9fans] drawterm

[Gabriel Diaz]

[-- Attachment #1: Type: text/plain, Size: 1257 bytes --]

Hi

I just downloaded the windows version and seems that this
new version can't be used with the same params of the old one.

I try to run it without params and with the common /? -h --help,
and i only get a crash. I think it try to open some environment
variables that are missing. (i can send you the complete
dump from windows if needed)

How can be used this new version? are there some env vars I must
use?

Thanks

Gabi


2005/11/4, Russ Cox <rsc@swtch.com>:
>
> I am almost ready to declare dt2k "stable".
>
> The random lock-ups were due to suspect X11 use
> that I cleaned up a few weeks ago.
>
> The fcp bug I mentioned before turns out to be a
> problem with Plan 9, not with drawterm. The bug was
> in devssl (or in devmnt, depending on how you look at it),
> not in drawterm. There is a new /sys/src/9/port/devssl.c
> and devmnt.c on sources.
>
> There are new drawterm binaries at http://swtch.com/drawterm/
>
> If you build yourself a kernel with the new devssl/devmnt
> and are running the latest drawterm and see it crash or
> do something else wrong, please let me know.
>
> The latest drawterm has one new feature: if you are running
> a plan9port factotum, then drawterm will use it.
>
> Russ
>

[-- Attachment #2: Type: text/html, Size: 1644 bytes --]

Re: [9fans] drawterm

[Axel Belinfante]

> There are new drawterm binaries at http://swtch.com/drawterm/

I almost don't dare to ask...
anybody tried to compile it (with success)
for sunos 5.8?

Axel.

Re: [9fans] drawterm

[Russ Cox]

>  I try to run it without params and with the common /? -h --help,
>  and i only get a crash. I think it try to open some environment
>  variables that are missing. (i can send you the complete
>  dump from windows if needed)

When I run from unix with no params, a drawterm window pops up
that says "set $cpu" and sits there.  That's what's supposed to
happen in Windows.  I don't know why it doesn't.

The command line is

  drawterm [-c cpuserver] [-a authserver] [-s secstoreserver]

and some other flags which don't matter much.  If you use -a
but not -s, -s will be filled in with the arg from -a.

If you don't specify -c it uses $cpu.  You have to specify -a unless
you're using factotum (Unix only).

Russ

Re: [9fans] drawterm

[Gabriel Diaz]

[-- Attachment #1: Type: text/plain, Size: 1032 bytes --]

Hi

That were the flags i used to use with the old one.
I run it in windows xp.

I tryed the binary Nils put and it run fine..
With env vars and with the usual -a and -c parameters works.
(and -s param too )

thanks

Gabi



2005/11/4, "Nils O. Selåsdal" <noselasd@asgaard.homelinux.org>:
>
> Gabriel Diaz wrote:
> >
> > Hi
> >
> > I just downloaded the windows version and seems that this
> > new version can't be used with the same params of the old one.
> >
> > I try to run it without params and with the common /? -h --help,
> > and i only get a crash. I think it try to open some environment
> > variables that are missing. (i can send you the complete
> > dump from windows if needed)
> http://asgaard.homelinux.org/plan9/dt2k-mingw.exe doesn't crash atleast
> if given /?, -h or --help
>
> > How can be used this new version? are there some env vars I must
> > use?
>
> run it with the flags
> -a the.auth.server -c the.cpu.server
> (-u user and -s secstore can be specified as well)
>

[-- Attachment #2: Type: text/html, Size: 1453 bytes --]

Re: [9fans] drawterm

[Axel Belinfante]

> > There are new drawterm binaries at http://swtch.com/drawterm/
> 
> I almost don't dare to ask...
> anybody tried to compile it (with success)
> for sunos 5.8?

To answer my own question: after a little struggle I have a drawterm.

When I start it with my usual -a and -c flags it says:

| cpu: cannot write user/domain choice in p9any:
| '/mnt/factotum/ctl' does not exist
| 
| goodbye

Is this something I missed in the struggle?
Or the effect of a too old p9p installation?
Or?

(running cpu from a plan9 terminal works fine,
 as does old drawterm) 

Axel.

Re: [9fans] drawterm

[Axel Belinfante]

In your message to me you wrote:
In message <200511042214.jA4MEGX17643@demeter.cs.utwente.nl>you write:
> > > There are new drawterm binaries at http://swtch.com/drawterm/
> > 
> > I almost don't dare to ask...
> > anybody tried to compile it (with success)
> > for sunos 5.8?
> 
> To answer my own question: after a little struggle I have a drawterm.
> 
> When I start it with my usual -a and -c flags it says:
> 
> | cpu: cannot write user/domain choice in p9any:
> | '/mnt/factotum/ctl' does not exist
> | 
> | goodbye
> 
> Is this something I missed in the struggle?
> Or the effect of a too old p9p installation?
> Or?
> 
> (running cpu from a plan9 terminal works fine,
>  as does old drawterm) 

commenting out the 'exits' at the end of 'fatal'
and adding some debug prints shows me a bit more.
seems like the '/mnt/factotum/ctl does not exist
somehow is stuck from a previous error and not reset.

reading p9any negotiation: p9sk1@cs.utwente.nl
cpu: cannot write user/domain choice in p9any: 11 != 20: p9sk1 cs.utwente.nl
cpu: cannot write p9sk1 challenge: (factotum ctl err msg)
user@cs.utwente.nl password:
cpu: cannot send ticket and authenticator back in p9sk1: cs gave empty translation list
cpu: can't authenticate slurp: cs gave empty translation list
cpu: wrting network: dir: fd out of range or not open
cpu: wanting for FS: fd out of range or not open: fd out of range or not open
remote cpu:
goodbye


Axel.

Re: [9fans] drawterm

[Brantley Coile]

[-- Attachment #1: Type: text/plain, Size: 127 bytes --]

i just downloaded it.  ran it under fc4 and it drops core.
now, if i can just find out where they put the core file!!

 bc

[-- Attachment #2: Type: message/rfc822, Size: 4317 bytes --]

From: Axel Belinfante <Axel.Belinfante@cs.utwente.nl>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] drawterm
Date: Fri, 04 Nov 2005 23:14:16 +0100
Message-ID: <200511042214.jA4MEGX17643@demeter.cs.utwente.nl>

> > There are new drawterm binaries at http://swtch.com/drawterm/
> 
> I almost don't dare to ask...
> anybody tried to compile it (with success)
> for sunos 5.8?

To answer my own question: after a little struggle I have a drawterm.

When I start it with my usual -a and -c flags it says:

| cpu: cannot write user/domain choice in p9any:
| '/mnt/factotum/ctl' does not exist
| 
| goodbye

Is this something I missed in the struggle?
Or the effect of a too old p9p installation?
Or?

(running cpu from a plan9 terminal works fine,
 as does old drawterm) 

Axel.

Re: [9fans] drawterm

[andrey mirtchovski]

run 'ulimit -c unlimited' on the shell and you'll get core files.

i just compiled it on fc4 and got a segfault:

#0  0x0804c934 in secdial (secstore=0x0) at secstore.c:35
35              if(*p == 0)       /* else use the authserver */


On Fri, 4 Nov 2005, Brantley Coile wrote:

> i just downloaded it.  ran it under fc4 and it drops core.
> now, if i can just find out where they put the core file!!
>
> bc
>

Re: [9fans] drawterm

[Brantley Coile]

[-- Attachment #1: Type: text/plain, Size: 104 bytes --]

i did the ulimit thing.  i just forgot what the core file was named.
i see core.<pid> now.

thanks.

[-- Attachment #2: Type: message/rfc822, Size: 3264 bytes --]

From: andrey mirtchovski <andrey@lanl.gov>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] drawterm
Date: Fri, 4 Nov 2005 16:01:45 -0700 (MST)
Message-ID: <Pine.LNX.4.64.0511041600510.23854@mtdew.lanl.gov>

run 'ulimit -c unlimited' on the shell and you'll get core files.

i just compiled it on fc4 and got a segfault:

#0  0x0804c934 in secdial (secstore=0x0) at secstore.c:35
35              if(*p == 0)       /* else use the authserver */


On Fri, 4 Nov 2005, Brantley Coile wrote:

> i just downloaded it.  ran it under fc4 and it drops core.
> now, if i can just find out where they put the core file!!
>
> bc
>

Re: [9fans] drawterm

[Tim Wiess]

> i just downloaded it.  ran it under fc4 and it drops core.
> now, if i can just find out where they put the core file!!

I noticed that it seg faults when you don't specify an auth server. Due
to a null dereference in netmkaddr(authserver, "tcp", "567") from
authdial().

I added the patch below to fix it as well as provide support for $auth.


Index: cpu.c
===================================================================
RCS file: /cvs/drawterm/cpu.c,v
retrieving revision 1.3
diff -u -r1.3 cpu.c
--- cpu.c	4 Nov 2005 17:06:57 -0000	1.3
+++ cpu.c	4 Nov 2005 22:59:48 -0000
@@ -181,6 +181,12 @@
 			fatal(0, "set $cpu");
 		system = p;
 	}
+	if(authserver == nil) {
+		p = getenv("auth");
+		if(p == 0)
+			fatal(0, "set $auth");
+		authserver = p;
+	}
 
 	if(err = rexcall(&data, system, srvname))
 		fatal(1, "%s: %s", err, system);

Re: [9fans] drawterm

[andrey mirtchovski]

it looks like it has always been there. i triggered it only when i ran
drawterm with bogus parameters.

obviously i don't make a good tester as i switched to it exclusively
several months ago :)

Re: [9fans] drawterm on solaris

[Axel Belinfante]

> > > > There are new drawterm binaries at http://swtch.com/drawterm/
> > > 
> > > I almost don't dare to ask...
> > > anybody tried to compile it (with success)
> > > for sunos 5.8?
> > 
> > To answer my own question: after a little struggle I have a drawterm.
> > 
> > When I start it with my usual -a and -c flags it says:
> > 
> > | cpu: cannot write user/domain choice in p9any:
> > | '/mnt/factotum/ctl' does not exist
> > | 
> > | goodbye
> > 
> > Is this something I missed in the struggle?
> > Or the effect of a too old p9p installation?
> > Or?
> > 
> > (running cpu from a plan9 terminal works fine,
> >  as does old drawterm) 
> 
> commenting out the 'exits' at the end of 'fatal'
> and adding some debug prints shows me a bit more.
> seems like the '/mnt/factotum/ctl does not exist
> somehow is stuck from a previous error and not reset.
> 
> reading p9any negotiation: p9sk1@cs.utwente.nl
> cpu: cannot write user/domain choice in p9any: 11 != 20: p9sk1 cs.utwente.nl
> cpu: cannot write p9sk1 challenge: (factotum ctl err msg)
> user@cs.utwente.nl password:
> cpu: cannot send ticket and authenticator back in p9sk1: cs gave empty translation 
> list
> cpu: can't authenticate slurp: cs gave empty translation list
> cpu: wrting network: dir: fd out of range or not open
> cpu: wanting for FS: fd out of range or not open: fd out of range or not open
> remote cpu:
> goodbye

If solaris is the only system where the following breaks,
seems we know again why we still love it, even if just a bit -
to keep us honest.

(the above is a probably bad attempt at paraphrasing the Geoff's
  'active port [...] to keep the code honest' line at the wiki)

anyway, the fix below brings me to:
 'can't authenticate: cpuserver: can't establish ssl connection: bad algorithm'

I'll continue digging.

Axel.

; cvs diff -buw  devip-posix.c
Index: devip-posix.c
===================================================================
RCS file: /cvs/drawterm/kern/devip-posix.c,v
retrieving revision 1.1
diff -b -u -w -r1.1 devip-posix.c
--- devip-posix.c       8 Aug 2005 12:50:11 -0000       1.1
+++ devip-posix.c       5 Nov 2005 21:26:36 -0000
@@ -199,7 +199,7 @@
 int
 so_send(int fd, void *d, int n, int f)
 {
-       send(fd, d, n, f);
+       return send(fd, d, n, f);
 }
 
 int

Re: [9fans] drawterm on solaris

[Axel Belinfante]

I'm sorry for the noise, nevertheless, after posting I thought:
I used the solaris compiler instead of gcc,
maybe that explains why I stumbled over this -
I could imagine all other compiles have been using gcc?

Axel.

> If solaris is the only system where the following breaks,
> seems we know again why we still love it, even if just a bit -
> to keep us honest.
> 
> (the above is a probably bad attempt at paraphrasing the Geoff's
>   'active port [...] to keep the code honest' line at the wiki)
> 
> anyway, the fix below brings me to:
>  'can't authenticate: cpuserver: can't establish ssl connection: bad algorithm'
> 
> I'll continue digging.
> 
> Axel.
> 
> ; cvs diff -buw  devip-posix.c
> Index: devip-posix.c
> ===================================================================
> RCS file: /cvs/drawterm/kern/devip-posix.c,v
> retrieving revision 1.1
> diff -b -u -w -r1.1 devip-posix.c
> --- devip-posix.c       8 Aug 2005 12:50:11 -0000       1.1
> +++ devip-posix.c       5 Nov 2005 21:26:36 -0000
> @@ -199,7 +199,7 @@
>  int
>  so_send(int fd, void *d, int n, int f)
>  {
> -       send(fd, d, n, f);
> +       return send(fd, d, n, f);
>  }
>  
>  int

Re: [9fans] drawterm on solaris

[Axel Belinfante]

> anyway, the fix below brings me to:
>  'can't authenticate: cpuserver: can't establish ssl connection: bad algorithm'

My bad - i was using a -e command line arg instead of intended -s.

It seems to work on solaris 5.8.

I compiled it with sun cc (grepped out all the //comments for it)
I added a posix-sun4u directory and populated it with files from
posix-power and posix-386, tas.s I took from p9p.

Axel.

[9fans] Re: drawterm

[Russ Cox]

I have fixed all the bugs that people reported.
The changes are in CVS and new binaries are at
http://swtch.com/drawterm/.  I believe the Windows
binary should work this time.

If anyone knows how to get mingw to set the
Windows binary's icon, please let me know.

Russ