diff options
| author | Craig Jennings <c@cjennings.net> | 2026-07-31 12:39:27 -0500 |
|---|---|---|
| committer | Craig Jennings <c@cjennings.net> | 2026-07-31 12:39:27 -0500 |
| commit | 0b93863689038a385fd320440ac0a413ce81c4e1 (patch) | |
| tree | 30dbabae88dfe704236544e5ebb6f6ebd18f4814 /tests/test-system-lib-confirm-destructive.el | |
| parent | 445a0be315f3d6fff585f1558262fb5dacd10d12 (diff) | |
| download | dotemacs-main.tar.gz dotemacs-main.zip | |
I had to kill an Emacs session today over a prompt I could not answer. A
second agent session held the selected window while "Overwrite hiroshi.m3u?
(yes or no)" waited in another frame, so my keystrokes went to a terminal
and the minibuffer stayed empty. Losing unsaved buffers is a far worse
outcome than a mis-keyed confirm.
So cj/confirm-strong becomes cj/confirm-destructive, and instead of binding
use-short-answers to nil for a typed "yes" it reads a single y or n.
Two things I kept:
- No default. Only y and n answer, so a stray RET or space re-prompts
rather than confirming a shutdown.
- Pending input is discarded before the read. Without that the change would
have been a bad trade: read-char-choice reads the input queue, so a key
typed before the prompt painted would confirm instantly, where the old
typed-"yes" absorbed it harmlessly.
Worth being honest about what this does not fix. One keystroke does not
make a prompt reachable when focus is elsewhere, and C-g is still the
escape. It lowers the cost of the situation rather than preventing it.
Six call sites: shutdown and reboot, shred, playlist overwrite and delete,
and the two file overwrites.
Diffstat (limited to 'tests/test-system-lib-confirm-destructive.el')
| -rw-r--r-- | tests/test-system-lib-confirm-destructive.el | 141 |
1 files changed, 141 insertions, 0 deletions
diff --git a/tests/test-system-lib-confirm-destructive.el b/tests/test-system-lib-confirm-destructive.el new file mode 100644 index 00000000..4fc96e21 --- /dev/null +++ b/tests/test-system-lib-confirm-destructive.el @@ -0,0 +1,141 @@ +;;; test-system-lib-confirm-destructive.el --- Tests for cj/confirm-destructive -*- lexical-binding: t; -*- + +;;; Commentary: +;; ERT tests for `cj/confirm-destructive', the confirmation used for +;; irreversible actions: file destruction, overwrites, power-off. +;; +;; The contract has two halves, and they pull against each other: +;; +;; 1. One keystroke. This replaced a typed-"yes" prompt on 2026-07-31, after +;; one became unanswerable -- a second agent session held the selected +;; window while the prompt waited in another frame, so keystrokes went to a +;; terminal and the Emacs session had to be killed with buffers unsaved. A +;; long-form prompt is only as safe as it is answerable. +;; +;; 2. No default. Only y and n answer. A stray RET or space re-prompts +;; instead of confirming, which is the accidental-confirm protection the +;; long form existed for, and it survives the change. + +;;; Code: + +(require 'ert) +(require 'cl-lib) +(require 'system-lib) + +;;; Normal Cases + +(ert-deftest test-system-lib-confirm-destructive-returns-t-on-y () + "Normal: a y answer confirms." + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?y))) + (should (eq (cj/confirm-destructive "Really? ") t)))) + +(ert-deftest test-system-lib-confirm-destructive-returns-nil-on-n () + "Normal: an n answer declines." + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?n))) + (should (eq (cj/confirm-destructive "Really? ") nil)))) + +;;; Boundary Cases + +(ert-deftest test-system-lib-confirm-destructive-accepts-uppercase () + "Boundary: a capital Y confirms and a capital N declines, so the answer +does not depend on the shift key or caps lock." + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?Y))) + (should (eq (cj/confirm-destructive "Really? ") t))) + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?N))) + (should (eq (cj/confirm-destructive "Really? ") nil)))) + +(ert-deftest test-system-lib-confirm-destructive-offers-only-y-and-n () + "Boundary: RET, space and every other key are refused. + +This is the protection that justified the old typed-\"yes\" form, kept now +that the answer is a single keystroke. Asserted on the character set handed +to `read-char-choice', which is what actually decides, rather than on the +prompt text, which only describes." + (let (chars) + (cl-letf (((symbol-function 'read-char-choice) + (lambda (_prompt cs &rest _) (setq chars cs) ?n))) + (cj/confirm-destructive "Really? ")) + (should (equal (sort (copy-sequence chars) #'<) '(?N ?Y ?n ?y))) + (should-not (memq ?\r chars)) + (should-not (memq ?\n chars)) + (should-not (memq ?\s chars)))) + +(ert-deftest test-system-lib-confirm-destructive-is-single-keystroke () + "Boundary: it never routes through `yes-or-no-p'. + +The regression this file guards. Binding `use-short-answers' to nil for the +call, which is what the old implementation did, is exactly the shape that +produced an unanswerable prompt." + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?y)) + ((symbol-function 'yes-or-no-p) + (lambda (&rest _) (error "must not demand a typed yes")))) + (should (eq (cj/confirm-destructive "Really? ") t)))) + +(ert-deftest test-system-lib-confirm-destructive-ignores-short-answer-setting () + "Boundary: the answer is one key whether or not `use-short-answers' is set. +The global default is t, but nothing about this prompt should depend on it." + (dolist (setting '(t nil)) + (let ((use-short-answers setting)) + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ?y)) + ((symbol-function 'yes-or-no-p) + (lambda (&rest _) (error "must not demand a typed yes")))) + (should (eq (cj/confirm-destructive "Really? ") t)))))) + +(ert-deftest test-system-lib-confirm-destructive-discards-type-ahead () + "Boundary: pending input is discarded before the key is read. + +The regression that made this necessary. `read-char-choice' reads the input +queue, so a key typed before the prompt appeared would answer it: a queued y +would confirm a shutdown or a file deletion instantly. The typed-\"yes\" +form absorbed such a key harmlessly, so dropping it without this guard would +have traded a rare annoyance for a rare catastrophe. + +Asserted on ordering, since that is the whole property: the discard has to +happen before the read, not merely somewhere in the function." + (let ((events '())) + (cl-letf (((symbol-function 'discard-input) + (lambda (&rest _) (push 'discard events) nil)) + ((symbol-function 'read-char-choice) + (lambda (&rest _) (push 'read events) ?y))) + (should (eq (cj/confirm-destructive "Really? ") t))) + (should (equal (nreverse events) '(discard read))))) + +;;; Error Cases + +(ert-deftest test-system-lib-confirm-destructive-declines-on-other-char () + "Error: anything that is not y or n declines rather than confirming. + +`read-char-choice' normally loops until it gets a listed character, but it +can return outside the set -- `read-char-from-minibuffer' yields RET when the +minibuffer result is empty. On an irreversible action the safe reading of an +unexpected answer is no." + (dolist (ch (list ?\r ?\n ?\s ?q ?\C-m)) + (cl-letf (((symbol-function 'read-char-choice) (lambda (&rest _) ch))) + (should-not (cj/confirm-destructive "Really? "))))) + +(ert-deftest test-system-lib-confirm-destructive-propagates-quit () + "Error: C-g aborts the caller rather than reading as a confirmation. +Swallowing the quit here would turn an abort into a yes on an irreversible +action." + (cl-letf (((symbol-function 'read-char-choice) + (lambda (&rest _) (signal 'quit nil)))) + ;; `should-error' cannot express this: quit is not an error, so ERT lets + ;; it through and reports the test as QUIT rather than passed. Catching + ;; it here is what makes the assertion a real pass or fail. + (should (eq 'quit (condition-case nil + (progn (cj/confirm-destructive "Really? ") + 'returned-normally) + (quit 'quit)))))) + +(ert-deftest test-system-lib-confirm-destructive-prompt-shows-choices () + "Error: the prompt names the keys that answer, so an unfamiliar prompt is +not a guessing game." + (let (prompt) + (cl-letf (((symbol-function 'read-char-choice) + (lambda (p &rest _) (setq prompt p) ?n))) + (cj/confirm-destructive "Delete everything? ")) + (should (string-match-p "Delete everything\\?" prompt)) + (should (string-match-p "y or n" prompt)))) + +(provide 'test-system-lib-confirm-destructive) +;;; test-system-lib-confirm-destructive.el ends here |
