Bjørn Nordbø | 1 Sep 2002 12:00

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

Kai.Grossjohann <at> CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> No, no.  That's not what I meant.
> 
> You have to interactively log in to the remote host by doing

Ah, ok. Well, I tried that and it worked fine (started Emacs and a
shell buffer and yanked the uuencoded data into the standard in of
the uudecode-process. It seemed to work fine.

I have done some more testing:

- Have tested real.hosts being both Solaris 8 and Debian GNU/Linux;
  no difference.
- The problem seems to occur as the file goes above 1005 bytes. I
  have tested this several times; 1005 bytes is of, 1006 bytes and
  it hangs.
- I have found no runaway processes on any of the hosts, just a
  chain of [emacs -> ssh] -> [sshd -> bash -> ssh] -> [sshd -> sh]
  where everybody seems to be expecting some kind of response from
  somewhere (the last shell hangs on read from FD 0).

However there's one thing that troubles me. Even though I use the
"multi" method (which should use mimencode) and has set mimencode
as tramp-encoding-command and tramp-decoding-command, it seems from
the debug buffer that Tramp still uses uudecode. Hmm.

Thanks again for you answers.

Bjorn
(Continue reading)

Kai Großjohann | 1 Sep 2002 21:29
Picon
Picon

Version 2.0.15

I have now released version 2.0.15 with bug fix changes from 2.0.14.
Thanks to Michael Albinus and Steve Youngs for providing the bug
fixes.

The new version is not in Emacs, yet.  Maybe it's easier to do this
from my office account.

I wonder what about XEmacs?  Should the new version be integrated
into XEmacs, too?

kai
--

-- 
A large number of young women don't trust men with beards.  (BFBS Radio)
Kai Großjohann | 1 Sep 2002 18:45
Picon
Picon

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

bn <at> nextra.com (Bjørn Nordbø) writes:

> Kai.Grossjohann <at> CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>> No, no.  That's not what I meant.
>> 
>> You have to interactively log in to the remote host by doing
>
> Ah, ok. Well, I tried that and it worked fine (started Emacs and a
> shell buffer and yanked the uuencoded data into the standard in of
> the uudecode-process. It seemed to work fine.

Hm.  So why does it fail when Tramp tries to do it?  Hm.

> I have done some more testing:
>
> - Have tested real.hosts being both Solaris 8 and Debian GNU/Linux;
>   no difference.
> - The problem seems to occur as the file goes above 1005 bytes. I
>   have tested this several times; 1005 bytes is of, 1006 bytes and
>   it hangs.

Interesting that the problem is so hard -- no fuzziness involved, it
seems.  Does it work to send the data from within Tramp in increments
of, say, 512 bytes, and waiting a bit between chunks?

Can you frob the Tramp source code to do that and try?

> - I have found no runaway processes on any of the hosts, just a
>   chain of [emacs -> ssh] -> [sshd -> bash -> ssh] -> [sshd -> sh]
>   where everybody seems to be expecting some kind of response from
(Continue reading)

Istvan Marko | 3 Sep 2002 10:33

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

Bjørn Nordbø <bn <at> telenor.net> writes:

> I would like to edit files on a box on our internal network going through
> a bastion host. It seems to work fine for very small files, but for slightly
> larger files tramp hangs when I try to save the file. The limit seems to go
> very close to 1000 bytes.
>
> The connection looks like this:
>
> [laptop]  ->  [[bastion host]  ->  [login host]]  ->  [internal server]

Bjørn,

Do you happen to have any NetBSD systems in the loop? I have seen
similar symptoms with one of our NetBSD system acting as the
remote. 

--

-- 
	Istvan
Bjørn Nordbø | 3 Sep 2002 11:09

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

Istvan Marko <mi-tramp <at> imarko.dhs.org> writes:

> Bjørn Nordbø <bn <at> telenor.net> writes:
> 
> > I would like to edit files on a box on our internal network going through
> > a bastion host. It seems to work fine for very small files, but for slightly
> > larger files tramp hangs when I try to save the file. The limit seems to go
> > very close to 1000 bytes.
> >
> > The connection looks like this:
> >
> > [laptop]  ->  [[bastion host]  ->  [login host]]  ->  [internal server]
> 
> Do you happen to have any NetBSD systems in the loop? I have seen
> similar symptoms with one of our NetBSD system acting as the
> remote. 

No, my laptop is running Debian GNU/Linux unstable, the bastion host
is running Solaris (8, I assume), the login host is running Solaris
8 I have tried using target hosts running both Solaris 8 and Debian
GNU/Linux. The Solaris hosts should have /local/gnu/bin first in their
PATHs.

Thanks anyway. :-)

--

-- 
I've seen things you people wouldn't believe. Attack ships on fire off
the shores of the Nonestic Ocean. I watched magic blunderbusses flash
and glitter in the dark near Glindea's Palace. All those moments will
be lost in time, like tears in rain. Time to die...
(Continue reading)

Kai Großjohann | 3 Sep 2002 14:52
Picon
Picon

XEmacs: does dired still work?

Volunteers needed to test the current CVS version on XEmacs.  Does
dired still work?

Thanks,
kai
--

-- 
A large number of young women don't trust men with beards.  (BFBS Radio)
Kai Großjohann | 5 Sep 2002 18:17
Picon
Picon

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

Bjørn Nordbø <bn <at> telenor.net> writes:

> I would like to edit files on a box on our internal network going through
> a bastion host. It seems to work fine for very small files, but for slightly
> larger files tramp hangs when I try to save the file. The limit seems to go
> very close to 1000 bytes.

I have no idea what might be the problem, but I have an idea that
maybe it's possible to work around it.  I have written the following
patch which allows you to send the file to the remote host "slower".

If you do (setq tramp-chunksize 500), then files will be sent to the
remote host in pieces of 500 bytes and Tramp will wait for 0.1
seconds after each chunk before sending the next one.

Does this help?  Maybe you could try different values for
tramp-chunksize.  Maybe you could also try to replace (sleep-for 0.1)
there with (sleep-for 1) or so.  The number is the number of seconds
to wait between chunks.

cvs server: Diffing .
cvs server: Diffing contrib
cvs server: Diffing lisp
Index: lisp/ChangeLog
===================================================================
RCS file: /cvsroot/tramp/tramp/lisp/ChangeLog,v
retrieving revision 2.183
diff -u -r2.183 ChangeLog
--- lisp/ChangeLog	3 Sep 2002 12:50:21 -0000	2.183
+++ lisp/ChangeLog	5 Sep 2002 16:14:20 -0000
(Continue reading)

Kai Großjohann | 5 Sep 2002 18:48
Picon
Picon

Re: Tramp and Zope.

Kai.Grossjohann <at> cs.uni-dortmund.de (Kai Großjohann) writes:

> artur <at> artgabi.com.pl (Artur Maciąg) writes:
>
>>   Can I test it somehow ? 
>>   Error occurs with options -q --no-site-file, too.
>
> Well, the variable auto-save-file-name-transforms says to write
> auto-save files to a local file.  Now the question is which Tramp
> function gets invoked when auto-saving so that this Tramp function
> can get improved to do the right thing with these files.
>
> Ah, here's an idea: you change the function tramp-file-name-handler
> to print a message with the operation and the args, and then you do
> stuff normally, and then in the message output you look where was
> Tramp invoked from auto-save.  (You can find that spot by looking for
> the filename that you got in the previous backtrace.)
>
> I guess it will be obvious what to do as soon as this information is
> there.
>
> (defun tramp-file-name-handler (operation &rest args)
>   "Invoke tramp file name handler.
> Falls back to normal file name handler if no tramp file name handler exists."
>   (message "Tramp-fh: op %s args %s" operation args)
>   (let ((fn (assoc operation tramp-file-name-handler-alist)))
>     (if fn
> 	(catch 'tramp-forward-to-ange-ftp
>           (message "Tramp-call: op %s args %s" (cdr fn) args)
> 	  (save-match-data (apply (cdr fn) args)))
(Continue reading)

Artur Maciąg | 6 Sep 2002 08:26
Picon

Re: Tramp and Zope.

Kai.Grossjohann <at> CS.Uni-Dortmund.DE (Kai Großjohann) writes:

  [ cut ] 

> It's been a while since I sent this message, but I can't remember an
> answer.  Do you still have the problem?  Do you still want to
> investigate it?

  Sorry, I was very busy.
  Since 30.08.2002 (thread 'tramp and ange-ftp') everything works
 great. Thanks.

        Artur.

--

-- 
   Artur Maciąg                          http://www.artgabi.com.pl
----------------------------------------------------------------------
                        HARDWARE: n. The part you kick.
Bjørn Nordbø | 6 Sep 2002 09:38

Re: tramp (2.0.14); tramp hangs when storing files >1000 bytes

Kai.Grossjohann <at> CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> Bjørn Nordbø <bn <at> telenor.net> writes:
> > I would like to edit files on a box on our internal network going through
> > a bastion host. It seems to work fine for very small files, but for slightly
> > larger files tramp hangs when I try to save the file. The limit seems to go
> > very close to 1000 bytes.
> 
> I have no idea what might be the problem, but I have an idea that
> maybe it's possible to work around it.  I have written the following
> patch which allows you to send the file to the remote host "slower".

I'm amazed by the support you can get on Tramp, but unfortunately (not
really :), I'm on vacation now and won't have a chance to test it until
early next week. When I get back I'll test the patch and let you know
how it works.

Thank you for your help, and thanks to everybody for Tramp. It is prob-
ably the best thing since we started using ssh. :-)

--

-- 
I've seen things you people wouldn't believe. Attack ships on fire off
the shores of the Nonestic Ocean. I watched magic blunderbusses flash
and glitter in the dark near Glindea's Palace. All those moments will
be lost in time, like tears in rain. Time to die...

Gmane