Re: question about `tramp-file-name-handler'

[Thierry Volpiatto]

Hi Michael,

Michael Albinus <address@hidden> writes:

> Thierry Volpiatto <address@hidden> writes:
>
> Hi Thierry,
>
>> I have now an error because with this last change, nil is returned
>> instead of the <process> by the next call of `start-file-process'.
>
> Returning nil is the expected behaviour. Simply because no process has
> been started.
Hmm, well the process have been started from main "emacs" but refused
(catched 'suppress) by tramp because it came to early, right ?
I will see how to deal with this, it is anyway much better than running
locally, thanks.
 
>> I think that what should be returned is the new process but delayed in a
>> timer or something similar.
>
> `start-file-process' has not such a semantic that the process shall be
> delayed. If a process is respawned too early in a given time frame, it
> simply ceases to work.
Ok, should be ok.

>> Also, it seem it is hardcoded to 5s to throw or not `suppress', is there
>> a particular reason for 5s? or maybe you found by experience two
>> processes that run quickly under this delay fails?
>
> 5 sec is hard-coded, indeed. As I said earlier, this mechanism was not
> introduced for process starts. Instead, it is a defense mechanism when
> Tramp plays yo-yo. See the discussion in 
> <http://thread.gmane.org/gmane.emacs.tramp/7848/focus=7869>.
>
> Maybe we shall give it a defcustom, and maybe we shall allow this to be
> nil. Then you could set it to your convenience.
Yes would be good, or at least a defvar should do it.

Thanks.

-- 
  Thierry
Get my Gnupg key:
gpg --keyserver pgp.mit.edu --recv-keys 59F29997