Re: tramp (2.1.16-pre); Re: test for tramp-chunksize "does not work"

[David Abrahams]

On Apr 23, 2009, at 3:29 PM, Michael Albinus wrote:

> David Abrahams <address@hidden> writes:
>
> > changing
> >            (accept-process-output proc 1))
> > to
> >            (accept-process-output proc 0))
> >
> > made it complete, saying:
> >
> > Bytes sent: 50  Bytes received: 50
> > Bytes sent: 100 Bytes received: 100
> > Bytes sent: 150 Bytes received: 150
> > Bytes sent: 200 Bytes received: 200
> > Bytes sent: 250 Bytes received: 250
> > Bytes sent: 300 Bytes received: 300
> > You should set `tramp-chunksize' to a maximum of 300
>
> I don't believe we shall set the timeout to 0. It doesn't let too much
> bytes through, because just one chunk of accept-process-output can be
> read in 0 seconds.

Yeah, I did a few experiments and .01 seemed to give the same  
performance on my Mac, while generating much shorter traces ;-).

> In my case (Ubuntu 9.04), with the timeout of 1 second, I get
>
>  "You should set `tramp-chunksize' to a maximum of 6750"
>
> When I set the timeout to 0 seconds, I get
>
>  "You should set `tramp-chunksize' to a maximum of 250"

Oho.

> The second statement would be wrong for me. Therefore, I believe we
> shall keep the test as it is, knowing that there are serious  
> problems on
> MacOS these days, which are not because of Tramp.


So the question for me is, since I have added the attached advice to  
accept-process-output, should I also adjust tramp-chunksize to  
whatever the test tells me?

tramp-setup.el
Description: Binary data

--
David Abrahams
BoostPro Computing
http://boostpro.com