bug#60943: 28.2; ‘tramp-handle-file-regular-p’ misses symlinks

[Greg Pfeil <greg@technomadic.org>]

[-- Attachment #1: Type: text/plain, Size: 233 bytes --]

I think sockets (where the 0th char is ?s) are also considered regular files. Also, I think that patch checks whether the truename is a regular file, but I believe /all/ symlinks are considered regular.

How about this version?


[-- Attachment #2: tramp-file-regular-p.patch --]
[-- Type: application/octet-stream, Size: 584 bytes --]

diff --git a/lisp/tramp.el b/lisp/tramp.el
index 351def66..82fa2b61 100644
--- a/lisp/tramp.el
+++ b/lisp/tramp.el
@@ -4039,7 +4039,7 @@ Let-bind it when necessary.")
        ;; Sometimes, `file-attributes' does not return a proper value
        ;; even if `file-exists-p' does.
        (when-let ((attr (file-attributes filename)))
-	 (eq ?- (aref (file-attribute-modes attr) 0)))))
+	 (memq (aref (tramp-compat-file-attribute-modes attr) 0) '(?- ?l ?s)))))

 (defun tramp-handle-file-remote-p (filename &optional identification connected)
   "Like `file-remote-p' for Tramp files."

[-- Attachment #3: Type: text/plain, Size: 1366 bytes --]



It just changes the equality check to check for the set of valid values. If we had the mode as a number, we could just check the one bit for regular files (which I think is what ‘file-regular-p’ does).


> On Jan 19, 2023, at 06:12, Michael Albinus <michael.albinus@gmx.de> wrote:
> 
> Greg Pfeil <greg@technomadic.org> writes:
> 
> Hi Greg,
> 
>> I was having an issue where .dir-locals.el on remote machines weren’t getting picked up when they were a symlink. Local symlinks worked fine, and remote non-symlinks also worked.
>> 
>> Digging in, ‘tramp-handle-regular-file-p’ explicitly checks that the first character in ‘file-attribute-mode’ is ?-. However, for symlinks, that character will be ?l.
>> 
>> Modifying ‘dir-locals--all-files’ to check
>> 
>>  (or (file-regular-p f) (file-symlink-p f))
>> 
>> \(which is ostensibly redundant) instead of simply
>> 
>>  (file-regular-p f)
> 
> Indeed, there's a bug. Thanks for the report!
> 
>> fixes the behavior, since TRAMP will now check with both ‘tramp-handle-file-regular-p’ and ‘tramp-handle-file-symlink-p’. The correct fix is to make ‘tramp-handle-file-regular-p’ match the behavior of ‘file-regular-p’.
> 
> The appended patch fixes this. Eli, is it OK to push to the emacs-29 branch?
> 
> Best regards, Michael.
> 
> <Mail Attachment>