Message with inline ICS

[Chris00]

Expected or desired behavior

View the message

Actual behavior

The attached message (composed with Mu4e) triggers an error when displaying it with the Gnus mode (not the original one).
(I can send privately the original message which also starts the debugger when “Enter Debugger on Error” is set.)

Versions of mu, mu4e/emacs, operating system etc.

Emacs 26.3
Mu: commit 441f005

[Chris00]

Same with the latest Mu version (commit efd6710).

[Chris00]

The problem is with gnus-icalendar-with-decoded-handle: it does

  (let ((charset (make-symbol "charset")))
    `(let ((,charset (cdr (assoc 'charset (mm-handle-type ,handle)))))

and this returns an invalid charset.

[djcb]

Any idea how to fix this?

[Chris00]

On 25 March 2020 at 07:47 CET, Dirk-Jan C. Binnema ***@***.***> wrote: Any idea how to fix this?

I'll try to find some time to investigate in the coming days. I see two possibilities: one could try to add the charset information to the handle (we try to determine it in `mu4e~view-gnus`) or we could advise the function (using `mu4e~view-rendering` to only trigger it for mu4e).

[Chris00]

Temporary fix (a better solution must be found to use the charset in mu4e~view-gnus).

(define-advice gnus-icalendar-event-from-handle
    (:filter-args (handle-attendee) fix-missing-charset)
  "Fix mu not passing the charset."
  (let* ((handle (car handle-attendee))
         (attendee (cdr handle-attendee))
         (buf (mm-handle-buffer handle))
         (ty (mm-handle-type handle))
         (rest (cddr handle)))
    ;; Put the fallback at the end:
    (setq ty (append ty '((charset . "utf-8"))))
    (setq handle (cons buf (cons ty rest)))
    (list handle attendee)))

[djcb]

Ah, thanks. But should the advise perhaps have a guard so it only works when used in mu4e? Otherwise it could affect people trying to use the 'real' gnus after this is defined. Or?

[Chris00]

I filed a bug against Emacs. The following variant only apply the customization inside mu:

(define-advice gnus-icalendar-event-from-handle
    (:filter-args (handle-attendee) fix-missing-charset)
  "Fix mu not passing the charset."
  (if (boundp 'mu4e~view-rendering)
      (let* ((handle (car handle-attendee))
             (attendee (cdr handle-attendee))
             (buf (mm-handle-buffer handle))
             (ty (mm-handle-type handle))
             (rest (cddr handle)))
        ;; Put the fallback at the end:
        (setq ty (append ty '((charset . "utf-8"))))
        (setq handle (cons buf (cons ty rest)))
        (list handle attendee))
  handle-attendee))

[Chris00]

Emacs bug.

[djcb]

Ah, thanks again. But shouldn't the check be wether we're currently in an mu4e buffer? Otherwise the advise would still kick in if you use gnus after mu4e?

[Chris00]

One should replace (boundp 'mu4e~view-rendering)by (and (boundp 'mu4e~view-rendering) mu4e~view-rendering). mu4e~view-rendering may indeed be known outside mu4e buffers but is supposed to be set to t only in the code using Gnus view.

[djcb]

Ah, thanks. Do you want to make a PR, or do you think we can do this post-1.4?

[Chris00]

I'll try to make a PR later today — it will not affect many people but it is annoying when this bug is triggered.