On 11/10/2021 4:17 AM, Michael Albinus wrote:
> `tramp-find-foreign-file-name-handler' loops through
> `tramp-register-foreign-file-name-handler'. Only operations registered
> there need to support a VEC-OR-FILENAME argument.
> `tramp-archive-file-name-p' and `tramp-crypt-file-name-p' are not
> registered, so no change is needed.

Ah ha. I didn't realize they weren't included (I thought I just had to 
configure something to add them in 
`tramp-foreign-file-name-handler-alist'). In that case, it definitely 
makes sense eliminate the `accepts-vec' bits from my patch.

> We could (and should) inform the authors of both packages, that the
> signature for the `tramp-FOO-file-name-p' functions will change with
> Tramp 2.6. As I can see, there's no problem to adapt
> `magit-tramp-file-name-p' and `tramp-jumper-file-name-p'. And yes, an
> entry in etc/NEWS should be added as well.

Sounds good to me. I've added a NEWS entry that hopefully explains the 
change fairly clearly. Just to be extra-certain, I re-ran my performance 
tests from before and the results are the same.

> A  little bit more friendly for debugging:
> 
> --8<---------------cut here---------------start------------->8---
> 	;; The signature of `tramp-FOO-file-name-p' has changed, it
> 	;; expects a VEC here.
>          (when (with-demoted-errors "Error: %S" (funcall (car elt) vec))
> --8<---------------cut here---------------end--------------->8---

I went with this since it should make it easier for third parties to 
notice the issue and make the appropriate update to their code. Thanks 
for the pointer.