Re: tramp 2.1.11 completion list force refresh

[Thomas S. Chin]

Michael Albinus wrote:
> Yes, that's the scenario Tramp fails. It has cached the contents of
> "/ssh:remoteHost:/tmp/", and it doesn't see the new file "foo". It could
> detect it only when not caching, to the price of less performance.
>
> I don't know what is more important, in general. Opinions?

The most ideal is to have the best of both worlds.  I do understand it 
would be very complex to devise a scheme smart enough to cache and 
detect immediate changes to the underlying directory.  In most cases, 
this is not possible given the problem at hand.  I think it would be 
nice if there was a keystroke that could force a flush of the cache for 
that particular directory so a user can explicitly notify tramp that 
they want it to refresh the contents.

I admit that I am not intimately familiar with what options are at your 
disposal, but if it is possible to detect multiple consecutive tab 
keystrokes (e.g. 3), it could then refresh the cache entry for that 
particular directory.  If not, another explicit key binding that is 
consistent with other emacs bindings signifying "refresh" would be a 
second option.  Another option might be to have the cache layer refresh 
on every X accesses, so if a user were to hit tab X times, he is 
guaranteed to have it reload from the remote host.  You can make X a 
configurable parameter and default it to never reload if that is your 
preferred behavior.  A user like me that is willing to sacrifice some 
performance could change this setting to 1 if I always want it to 
reload, or some other number.

Tell me what you think about these options.

thom