fix(pdf-misc): do not signal from the size indication mode-line

[alberti42]

What happens

pdf-misc-size-indication is put into mode-line-position as an :eval construct, so it runs on every redisplay. When it signals, redisplay reports the error and drops the rest of the line:

Error during redisplay: (eval (pdf-misc-size-indication) t) signaled (wrong-type-argument overlayp nil)

Reverting a PDF while it is on screen — what AUCTeX does through TeX-revert-document-buffer when a recompile finishes — produced this two to six times per revert on a document with pdf-view-roll-minor-mode enabled.

Why

It calls pdf-view-image-size, which has nothing to measure whenever the page carries no image. With pdf-view-roll-minor-mode that function reads the overlay holding the page:

(overlay-get (pdf-roll-page-overlay page window) 'display)

so it signals rather than returning nothing. revert-buffer replaces the buffer text, and a page overlay carries evaporate, so the overlays collapse; until they are rebuilt every redisplay in that window reports the error.

The fix

Return an empty indication when the size is not available.

Guarding the read in pdf-view-image-size instead is not an option: some thirty-five callers across pdf-view.el, pdf-util.el, pdf-annot.el, pdf-links.el, pdf-sync.el and pdf-isearch.el do arithmetic on its result, so it cannot start returning nil. What that function should report when a page has no image yet is a separate question, and not one a mode-line construct needs answered — it just needs to not signal.

Normal reporting is unchanged: " Top" at the top of a page, " Bot" and the percentage below it.

Reproducing it

A one-shot advice that makes a revert land inside a page render. The lexical-binding cookie matters — without it the timer's lambda does not capture buf and win, and nothing happens. The revert also has to fire with a window other than the PDF's selected, which is what a revert from the TeX process sentinel does.

;; -*- lexical-binding: t; -*-
(defvar my-pdf-race-armed t)

(defun my-pdf-race (&rest _)
  (when my-pdf-race-armed
    (setq my-pdf-race-armed nil)
    (let* ((buf (current-buffer))
           (pdfwin (get-buffer-window buf t))
           (win (car (seq-remove (lambda (w) (eq w pdfwin))
                                 (window-list (window-frame pdfwin))))))
      (unless win (user-error "Split the PDF's frame first"))
      (run-at-time 0 nil
                   (lambda ()
                     (when (buffer-live-p buf)
                       (with-selected-window win
                         (with-current-buffer buf (revert-buffer t t)))))))))

(advice-add 'pdf-view-create-page :before #'my-pdf-race)

Open a PDF in a frame that has a second window, then M-x pdf-roll-goto-page to a page that has not been rendered yet, and read *Messages*. On a 16 page document the error appeared in two of two runs before this change and none of two after.

Priority

Lower than #361, #362 and #364. This is a cosmetic failure — noise in *Messages* and a briefly blank size indication — rather than a crash, and it corrects itself on the next redisplay.

It is also independent of all three: it touches only lisp/pdf-misc.el, and git merge-tree reports no conflict against any of them.