bug#58964: 28.2; `file-notify--handle-event` setting callback function to nil

[Michael Albinus]

Alejandro Pérez Carballo <apc@umass.edu> writes:

> Hi Michael, 

Hi Alejandro,

>> This is not a valid call of `file-notify--handle-event', see its
>> docstring how it shall be applied.
>
> I've looked at the docstring but don't really see why what I gave is
> not a valid call of the function.
>
>     Signature
>     (file-notify--handle-event DESC ACTIONS FILE FILE1-OR-COOKIE)
>     
>     Documentation
>     Handle an event returned from file notification.
>     
>     DESC is the back-end descriptor.  ACTIONS is a list of:
>      created
>      changed
>      attribute-changed
>      deleted
>      renamed           -- FILE is old name, FILE1-OR-COOKIE is new name or nil
>      renamed-from      -- FILE is old name, FILE1-OR-COOKIE is cookie or nil
>      renamed-to        -- FILE is new name, FILE1-OR-COOKIE is cookie or nil
>      stopped           -- no more events after this should be sent
>
> I thought the descriptor would just be one of the keys in the hash
> table that is the value of `file-notify-descriptors', but maybe I'm
> misunderstanding how that's all supposed to work. So basically, the
> call that's giving me the error is this:
>
>     (file-notify--handle-event 18 '(renamed attribute-changed deleted) 
> "/Users/apc/tmp/test.el" nil)
>
> Is that not a valid call? (I was passing as first argument `(car
> (hash-table-keys file-notify-descriptors))` because I thought it would
> be easier to use for reproduction purposes than asking people to see
> what the key associated with the visited file in
> `file-notify-descriptors` was.)

There are several problems. It is not clear whether desc 18 is related
to the file name you have given. The combination of actions you have
used is not valid, it makes no sense. When there is a renamed action,
the file1-or-cookie arg must be a file name. And I'm not sure whether we
want to see absolute file names.

Conclusio: Don't mess with this internal function.

> As for your other questions: I am not trying to call this function
> directly, but instead trying to find the source of an error message
> I've been getting on and off for months. After leaving
> `debug-on-error' on for a while, I finally got a backtrace. From what
> I could tell, that was the cause of the trouble. Something in my
> system is making a call to `file-notify--handle-event' with similar
> arguments (it was a different file, and a different key, the one that
> in fact was associated with the file in question in the
> `file-notify-descriptors` table).

For debugging purposes, I recommend to set file-notify-debug to t. This
shall give you a lot of information.

> Unfortunately, I cannot find the original backtrace information. If
> indeed what I provided was not a valid call of the function, I'll have
> to wait until next time the error shows up and see if I get more
> information about what could be making that invalid call in my system.

Yes, please. And if you still cannot determine what's up, pls contact
us. I'm closing *this* bug, because the problem you've reported doesn't
exist as such.

> Best, 
>
> A.

Best regards, Michael.