David Karr <email@example.com> writes:
> So, it seems like there's something about my process of creating shell
> buffers, which if done before the first tramp connection, makes tramp
> time out. I don't know what information tramp caches. It's possible
> that in the last scenario, if I then tried a tramp connection to a
> different host, that would fail, but I'm just guessing.
OK, that seems to be it.
> I create shell buffers using a couple of very small packages that I
> wrote decades ago, which I haven't had any issues with since. They
> manage a "ring" of shell buffers which I can either step through, or
> search for a particular one by its current directory. I'll attach
> them, if there's a clue there. Reading the file comment for
> "cycle-shell.el", I think I must have simplified it slightly after I
> wrote that comment, because I don't depend on a "shell-for-cycle"
> package anymore.
I've skimmed only shortly over the packages. However, you seem to create
shell buffers premature, before calling `shell'. So that function
expects, that the shell buffer is already prepared properly. But it
isn't I guess, `shell' does something more when creating a new buffer.
So I recommend to let `shell' do everything it believes what should be
done creating a buffer. You might debug your functions to see where they
block. And for your functionality, adding shell buffers to a ring, you
could add yourself into `shell-mode-hook', which seems to be the proper
I was thinking that I haven't really looked at this code for a very long time and might not be able to figure out how to make this happen, but I decided to give it a try, and I noticed in the doc for the "shell" function that it can take a prefix argument for the name of the buffer to use, instead of the default. That might be a better way to do this that might make it more compatible with tramp. I'll see if I can figure this out.
Before I noticed this, I had tried sending a request for help on this to gnu.emacs.help, but I got crickets.
Unfortunately, my first stab at this didn't help. I think the new function is an improvement, but it didn't fix the problem. Here's the new defun:
(defun cycle-make-shell ()
"Implements a ring of *shell* buffers. If current buffer is not a shell
buffer, then it will just execute 'shell'. If it IS a shell buffer, then
create a new shell buffer, starting with '*shell-1*', but skipping to the next
number if that already exists."
(let* ((bufname (buffer-name (current-buffer))))
(if (string-match "\*shell\\(\*\\|-\\([0-9][0-9]*\\)\*\\)" bufname)
(setq change-dir default-directory)
(setq done nil)
(while (not done)
(setq new-bufname (next-bufname bufname "shell"))
(if (bufferp (get-buffer new-bufname))
(setq bufname new-bufname)
(setq done t)
;; check for existence of buffer last-shell-buffer. If it exists,
;; go to it. If not, then execute "shell".
(if (and (bufferp last-shell-buffer)
(not (killed-buffer-p last-shell-buffer)))
Sorry that I cannot say too much about. Starting by tomorrow, I'll be
almost offline for some days.
Best regards, Michael.