Re: Reliably sending binary codes to TRAMP processes

JD Smith <[email protected]> Sat, 1 Feb 2025 14:47:54 -0500
Newsgroups gmane.emacs.tramp
Message-ID <[email protected]>

> On Feb 1, 2025, at 12:39 PM, Michael Albinus <[email protected]> wrote:
> 
> Michael Albinus <[email protected]> writes:
> 
> Hi,
> 
>>> I did eventually find this in the manual, but far away from where I
>>> looked (in fact I used consult-info).  It would be very helpful if the
>>> docstring for tamp-connection-properties method that some properties
>>> are mapped in this way.
>> 
>> Agreed. Will do.
> 
> I've added the link to the Tramp manual to the docstring. Pushed to the repositories.


Thank you.  I adopted as well your voodoo method to avoid polluting TRAMP's cache, and it seems to work well.  I've also discovered you can use a pipe connection to the direct-async SSH command, and it is very fast.

This is what you gain by "losing" EOF and real signal handling, being forced to send those over the wire. For now I have worked around TRAMP's notion that SIGINT must kill a process by writing my own trivial interrupt function.