| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
I copied the outreach email I sent on 2026-04-29 into assets/ so the maintainership-adoption thread has a tracked record in the repo. The .eml file follows the YYYY-MM-DD-<topic>-<descriptor> naming so future correspondence lands alongside it.
The email offered three options (push rights, formal handover, consolidated MR batch) and hasn't received a reply as of today, 29 days out.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I moved both v0 design specs out of working/ and into docs/design/. That's the conventional permanent home for project documentation, where engineers will look during implementation. working/ is meant for transient in-progress artifacts that file away once the work ships, and these specs are long-lived design docs that don't fit that contract.
Files moved:
- working/stats-dashboard/stats-dashboard.org → docs/design/stats-dashboard.org
- working/fsrs-spec/fsrs-spec.org → docs/design/fsrs-spec.org
The git rename detection picked both up, so file history follows the move.
I also dropped the stale /docs entry from .gitignore. The Makefile doesn't write to docs/ and nothing else references it as a build output, so the ignore was inherited cruft that would have silently dropped any tracked file under docs/.
I updated path references in seven spots: three docstring/comment refs in org-drill.el, one in tests/test-org-drill-session-record.el (the Commentary block), and three inside the specs themselves. Two refs in fsrs-spec.org now point at the correct location for its defcustom docstring and option description. One in stats-dashboard.org's References section points at the sister spec.
Full make test-unit green. eask compile clean.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I added the persist + recording layer that the future stats dashboard reads, plus the v0 design spec at working/stats-dashboard/stats-dashboard.org that scoped the work. Every completed (non-suspended) drill session now appends one org-drill-session-record to a persisted org-drill-session-log via persist-defvar, mirroring the SM5-matrix pattern that's already in place. A new sibling defgroup org-drill-statistics groups the dashboard's customs.
The record carries: start-time, end-time, scope, algorithm, qualities (as a vector), pass-percent, new-count, mature-count, failed-count, cram-mode. Scope and algorithm are captured at session-start time onto two new org-drill-session slots (scope-at-start, algorithm-at-start) so a mid-session defcustom flip doesn't misrepresent what was actually drilled. Both org-drill--prepare-fresh-session and the org-drill-again resume path populate the slots.
I extracted pass-percent into org-drill--compute-pass-percent and call it from both org-drill-final-report and org-drill-session-record-from-session, so the user-visible report and the persisted record can't drift on the rounding or the failure-quality threshold.
I wrapped recording in condition-case at the call site, not ignore-errors. Any recorder bug or persist-save IO failure gets messaged so silent data loss leaves a forensic trail.
Corrupt-load recovery follows the SM5 path's condition-case fallback and adds an org-drill--session-log-quarantine helper that renames the bad file to a .corrupt-YYYY-MM-DDTHHMMSS sibling so the next save doesn't overwrite it. The seconds-granularity suffix prevents a same-day double corruption from clobbering the earlier quarantine. The helper depends on persist--file-location, an internal symbol guarded behind fboundp and documented inline.
The spec at working/stats-dashboard/stats-dashboard.org is ratified, with all 10 originally-open decisions resolved in the Ratified Decisions table at the bottom: persist-defvar, warn-and-rename corrupt-load recovery, quadrant-block sparklines, a single buffer-wide filter, CSV with proper quoting, sync dashboard open, defer aborted-session recording, the q/g/e/s/r/a/RET keymap, leech-quality threshold default 2.5, and a sibling org-drill-statistics defcustom group. The remaining work (dashboard renderer, CSV export, docs) lands in follow-up commits.
I followed TDD throughout. 24 tests in tests/test-org-drill-session-record.el cover struct construction, the shared pass-percent helper, the builder including scope/algorithm capture and mature-count summing, log appending with newest-first ordering and persist-save call-through, the symbol-bound smoke check, the quarantine rename plus its seconds-granularity contract, and the end-message hook (records on normal completion, skips on suspend, logs to message on recorder error). Full make test-unit green. eask compile clean.
|
| | |
|
| |
|
|
|
|
|
|
| |
Closes upstream #46. I changed the default for org-drill-spaced-repetition-algorithm from sm5 to simple8. The defcustom now carries an ADR-style comment recording the reasoning: simple8 gives most of sm5's per-user adaptation value via its per-card difficulty learning, without sm5's dependency on a persisted optimal-factor matrix that can rot (the upstream #45 fragility), and it adjusts intervals for early or late reviews, a real-world concern sm2 and sm5 don't address.
Existing users with the option set in their config see no change. Existing users on the previous default get simple8 on their next session, with no state migration. Simple8 reads what it can from the SM-style item-data and carries on. Sm5's persisted optimal-factor matrix stays on disk and is available if the user switches back.
I added a test that pins the default value, so an accidental flip surfaces in CI.
|
| |
|
|
|
|
| |
I drafted a design spec for adding FSRS (Free Spaced Repetition Scheduler) as a fourth choice in org-drill-spaced-repetition-algorithm. It pins FSRS-4.5 as the version target, documents the DSR state model, the rating mapping from org-drill's 0-5 quality to FSRS's 1-4 scale, the update equations, the new DRILL_FSRS_* properties, and the integration shape into the existing scheduler dispatch.
The spec is v0 with six explicit DECIDE: markers carrying recommended defaults: version pin, parameter tuple, quality mapping, return shape, cold-start UI, and equation cross-check. The companion todo entry stays open at todo.org:125 until those land and implementation begins. The file moves to fsrs-spec.org at the project root once the DECIDEs resolve.
|
| |
|
|
|
|
|
|
| |
Stage 5 of #147, closing the scheduler migration. simple8 now takes (state quality &optional delta-days) instead of seven positional args, binding the recall fields from the struct at the top so the algorithm body is unchanged. simple8 doesn't use ease, so the binding skips that slot.
Both call-site branches collapse to (state quality [delta-days]), dropping the per-branch accessor unpacking. The testutil adapter test-scheduler--call-simple8 keeps the simple8 test calls a one-symbol rename per site. One direct simple8 call in tests/test-org-drill-prompt-and-format-helpers.el now uses the new struct API inline.
With this stage landed, all three schedulers, the item-data round-trip, and every test caller go through the org-drill-card-state struct, finishing #147.
|
| |
|
|
|
|
|
|
| |
Stage 4 of #147. sm5 now takes (state quality of-matrix &optional delta-days) instead of nine positional args, binding the recall fields from the struct at the top so the algorithm body is unchanged. Both call-site branches pass the state they already hold, dropping the per-branch accessor unpacking.
The testutil adapter test-scheduler--call-sm5 keeps the sm5 test calls a one-symbol rename per site. I also kept the return as the existing list, matching the stage-3 refinement: the goal is reducing the input signature, and changing the return shape would force the shared return-extractors and every return-read to change for no real gain.
Also folds in two stage-3 follow-ons I missed when sm2 landed: a direct sm5 call in tests/test-org-drill-small-branch-coverage.el now uses the new struct API inline, and five direct sm2 calls in the simple-workflow integration test now go through the testutil adapter (the integration file picks up the testutil-scheduler require). Caught by running make test-integration this stage, which I should have run on the sm2 stage.
|
| |
|
|
|
|
| |
Stage 3 of #147. sm2 now takes (state quality) instead of seven positional args, binding the recall fields from the struct at the top so the algorithm body is unchanged. The smart-reschedule and hypothetical-next-review-date call sites pass the state they already hold, which drops the per-branch accessor unpacking.
I kept the return as the existing positional list rather than restructuring it too. The goal is reducing the input signature, and changing the return shape would force the shared test extractors and every return-read to change for no real gain. A testutil adapter, test-scheduler--call-sm2, packs positional args into the struct, so the test call sites stay readable as the documented algorithm inputs and the migration is a one-symbol rename per call.
|
| |
|
|
| |
The card-type-system-complete integration test expected translate_number to be registered in org-drill-card-type-alist, but issue #43 removed that entry because its presenter never existed, and test-org-drill-translate-number-regression.el asserts it stays out. The integration test contradicted the shipped fix and had been failing on that assertion since the test suite landed. I dropped translate_number from its expected-types list and noted why inline.
|
| |
|
|
|
|
| |
Implements upstream #56. When org-drill-maximum-duration is reached, the session used to keep presenting the in-progress card and the again-queue until they drained, so the only way out was to finish them or interrupt. I added the defcustom org-drill-on-timeout-action with a discard-current value that ends the session as soon as the limit is hit, leaving the dropped cards untouched: they keep their existing scheduling and turn up again next session. The default, finish-current, preserves the old behavior.
The gate lives in org-drill-entries-pending-p, the single predicate the drill loop checks between cards. Under discard-current past the limit, the in-progress and again items stop counting as pending, so the loop ends instead of draining them.
|
| |
|
|
|
|
|
|
| |
Second step of #147. get-item-data now returns an org-drill-card-state and store-item-data takes one, so the six recall fields move as named slots instead of a positional list. The three call sites (smart-reschedule, hypothetical-next-review-date, copy-scheduling-to-marker) read scheduler inputs through accessors and build a struct for the store, which removes the hand re-ordering between the get-shape and the store-shape.
Behavior is unchanged. The legacy LEARN_DATA read path and the virgin-item sentinel are preserved field-for-field, and store takes just the struct because its last-interval slot already holds the interval to persist. The schedulers still take positional args; they adopt the struct in the following commits.
I updated the round-trip, integration, and setup-helper tests to build and read the struct via a small list-view helper, so the existing expected-value assertions stay readable.
|
| |
|
|
|
|
| |
First step of #147. I added a cl-defstruct bundling the recall fields the schedulers and the item-data round-trip shuttle around (last-interval, repetitions, ease, failures, meanq, total-repeats), with an ADR comment recording why: the same fields were passed as positional lists in three different orderings, and three call sites re-ordered between them by hand.
The struct is inert here. The item-data round-trip and the schedulers adopt it in the following commits.
|
| |
|
|
|
|
| |
I renamed several cryptic locals to say what they hold: idx to index-var in the pop-random gensym, val to raw-value in the five DRILL_* property getters, q to quality in hypothetical-next-review-dates, m to marker in free-markers, the a/b sort and filter lambdas in order-overdue-entries to entry/other, and dat to card-def in the empty-card-type check.
These are pure renames with no behavior change. Byte-compile and the full unit suite stay green.
|
| |
|
|
| |
sym1/sym2 gave no hint of their role. current-meter-char toggles per meter wrap and alternate-meter-char is its inverse, which is what the two make-string calls actually consume.
|
| |
|
|
| |
actions/checkout@v4 and emacs-eask/setup-eask@master both run on the node20 runtime GitHub force-migrates to node24 on 2026-06-02. checkout@v5 ships the node24 runtime, so I bumped it across all three jobs. setup-eask has no node24 release yet, so I set FORCE_JAVASCRIPT_ACTIONS_TO_NODE24 at the workflow level to run it under node24 now, ahead of the forced switch rather than after.
|
| |
|
|
|
|
| |
The three error-case tests in test-org-drill-determine-next-interval-sm2.el asserted their cl-assert preconditions with a bare should-error :type 'cl-assertion-failed. On Emacs 29, ERT installs a signal-hook that keeps should-error from matching the condition, so the tests failed the 29.4 job outright. Under undercover's edebug instrumentation the same assert dropped into a blocking debugger in batch mode, which hung the coverage job until the 6-hour timeout on every push.
The sm5 and simple8 files already wrap their equivalent error tests in a skip-unless guard for this quirk. I copied that helper into sm2 so all three scheduler files handle Emacs 29 the same way.
|
| |
|
|
| |
inbox/ holds local Claude handoff files and *~ are Emacs backups; neither belongs in the repo.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A batch of self-contained user-facing improvements, squashed from the feat/org-drill-solo-features branch.
I added an undo for the last rating (issue #2 follow-up). The rating prompt now takes an undo key (org-drill--undo-key, default u): it restores the previous card's scheduling snapshot, drops the recorded quality, and re-queues that card, then returns to the current prompt. Each rating snapshots the scheduling properties and SCHEDULED line onto a per-session stack capped at org-drill-undo-limit (default 3). org-drill-reschedule loops on the rating read so undo doesn't rate the current card.
I made the five session-control keys (quit, edit, help, skip, tags) defcustoms so they can be rebound from customize-group (issue #35), keeping their defaults. The 0-5 rating keys stay as-is, since they're tied to the quality scale rather than being variables.
I lifted the hardcoded 100-line entry-text limit in org-drill-get-entry-text into the org-drill-entry-text-max-lines defcustom, defaulting to 100.
I also deleted a commented-out old org-entry-empty-p that the real definition had already replaced.
Existing tests stay green and each change added its own, including snapshot/restore and prompt-loop tests for undo.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A cleanup pass over org-drill internals, squashed from the refactor/wave3-cleanup branch. No behavior change. Each step kept the existing tests green and added its own.
I shared two duplicated helpers across the language card getters: org-drill--read-property-string and org-drill--face-from-alist.
I factored the cloze body-scan out of the two multicloze presenters into org-drill--cloze-body-bounds, org-drill--count-cloze-matches, and org-drill--hide-cloze-by-index, so each presenter just picks which indices to hide.
I pulled the presenter resolution and the four-way result classification out of org-drill-entry-f into org-drill--resolve-presenter and org-drill--classify-presentation-result, untangling the pivot of every drill iteration.
I split the 37 defcustoms (and the three cloze faces) into four customize sub-groups (display, algorithm, session, leech) so customize-group org-drill is navigable. There's no leitner group because the Leitner settings are defvars.
I documented the 22 defuns that had no docstring, rewrote the corrupted org-drill-presentation-prompt-in-mini-buffer docstring, and switched eleven docstrings to the imperative "Return" (issue #2).
|
| |
|
|
|
|
|
|
| |
org-drill-add-cloze-fontification ran on org-font-lock-set-keywords-hook, which fires in every org buffer, and pushed the cloze rule into org's global org-font-lock-extra-keywords. The cloze regexp is built from the [ and ] delimiters, so an org priority cookie like [#A] matched the cloze pattern and got fontified as a cloze in every org buffer, colliding with org's headline fontification and stripping the heading's org-level-N face.
I replaced the global install with org-drill-mode, a buffer-local minor mode that adds the cloze keywords only to its own buffer via font-lock-add-keywords. org-drill-auto-enable-mode (default on) turns the mode on from org-mode-hook in buffers that hold drill cards, so existing drill files keep their cloze highlighting while plain org buffers stay clean. Highlighting still respects org-drill-use-visible-cloze-face-p.
The cloze regexp itself is unchanged, so the single-line cloze constraint from #38 is preserved.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A batch of test-coverage and hardening work, squashed from the test-work branch.
Tests: deduplicated a colliding leitner-capture test name so make test-name loads again. Added SM2 assert-failure cases, the six basic multicloze variant delegations, the three English-side spanish-verb branches, and org-drill-current-scope branch coverage.
Fix: the entry-point commands (org-drill itself, cram-tree, tree, directory, resume, relearn-item, strip-all-data, merge-buffers) carried no autoload cookies, so M-x failed from a fresh install until something pulled the file in. They're autoloaded now.
Perf: org-drill-shuffle was quadratic because it indexed a list with elt on every swap. It runs a linear Fisher-Yates pass over a vector now, and it checks its argument is a list.
Feat: added org-drill-version, a constant plus an interactive command, so a bug reporter doesn't have to open the file header.
Refactor: extracted org-drill--format-tense-mood, shared by the two verb-conjugation presenters that each carried a copy.
Docs: explained the SM8 magic numbers in the simple8 helpers as empirical fits rather than tunable knobs.
|
| |
|
|
|
|
| |
test-org-drill-entry-days-since-creation-with-date-added pinned DATE_ADDED to a hardcoded 2026-04-01 and asserted 33-35 days, which only holds when the wall clock is near 2026-05-05. The with-fixed-now mock that was supposed to make it deterministic rebinds `current-time`. But the DATE_ADDED branch of org-drill-entry-days-since-creation goes through `org-time-stamp-to-now`, which reads the real clock. The rebind never reaches it. CI went red once the calendar moved past the +35-day window.
I rewrote the test to derive DATE_ADDED 34 calendar days before today (via decode-time / encode-time, so month rollover and DST are handled) and assert the function returns exactly 34. No mock needed. Both sides read the same real clock.
|
| |
|
|
|
|
| |
The Cask→Eask migration in d012894 switched the Makefile to drive Eask, but ci.yml still installed Cask via cask/setup-cask. All three jobs (test, lint, coverage) failed at `make setup` with "Eask not found".
I swapped each job to use emacs-eask/setup-eask. No other CI logic changed.
|
| |
|
|
|
|
|
|
|
|
| |
Cask's upstream has slowed. Eask is the actively maintained successor.
Eask's `package-file` directive doesn't auto-install the deps from the .el header's Package-Requires, so the Eask file mirrors emacs/seq/org/persist explicitly. `eask install-deps` also doesn't pull transitive deps, so dash, m-buffer, and shut-up needed their own `depends-on` lines for undercover and elisp-lint to activate.
The Makefile swaps are mechanical: $(CASK) → $(EASK), `cask install` → `eask install-deps --dev`, `cask build` → `eask compile`.
The URL in org-drill.el's header pointed at the abandoned upstream's GitLab issues page. Eask cross-validates that against website-url, so I updated it to the GitHub mirror — where users file issues now.
|
| |
|
|
|
|
|
|
| |
I added a small case for `org-drill-response-get-buffer-create' that
sets an input method then asserts the helper propagates it into the
new response buffer.
Coverage stayed near 94.7% (instrumentation noise can move it ±0.2%).
|
| |
|
|
|
|
|
|
|
| |
I added a test for `org-drill-present-simple-card-with-typed-answer'
that mocks `prompt-for-string' and the surrounding hide-* helpers, then
asserts the presenter forwards the session and returns the prompt's
result.
Coverage moved from 94.4% to 94.8%.
|
| |
|
|
|
|
|
|
|
|
| |
I extended the presentation-prompt tests with cases for the in-buffer
variant: default prompt assembled when none is supplied, explicit
prompt fed straight through, drill-answer cleared on entry, and the
session's exit-kind flowing back to the caller after recursive-edit
returns. Recursive-edit, display-buffer, and the timer are all mocked.
Coverage moved from 93.2% to 94.4%.
|
| |
|
|
|
|
|
|
|
|
| |
I added a test for the developer helper `org-drill-test-display' that
mocks `org-drill-entry-f' to confirm the dispatcher fires and that the
zysygy tag is toggled off again on exit. I also added a small case
for `org-drill-test-display-rescheduler' that verifies it runs
`org-drill-display-answer-hook' and waits on read-key-sequence.
Coverage moved from 92.7% to 93.2%.
|
| |
|
|
|
|
|
|
|
| |
I extended the existing rating-key test with a dolist that fires every
remaining navigation key (up, left, right, prior, next) — previously
only `down' was covered. Each key advances the loop without ending it,
then a numeric input terminates.
Coverage moved from 92.3% to 92.7%.
|
| |
|
|
|
|
|
|
|
| |
I added a couple of cases for `org-drill-add-cloze-fontification':
when `org-drill-use-visible-cloze-face-p' is t, a fontification spec
gets pushed onto `org-font-lock-extra-keywords'; with the flag nil,
the list is left alone.
Coverage moved from 92.1% to 92.3%.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
goto-heading error
I added small-branch tests for the SM5 dispersal-factor multiplier
when `org-drill-add-random-noise-to-intervals-p' is t,
`org-drill--read-key-sequence' deactivating/reactivating an active
input method (and skipping the dance when none is active), and the
error branch in `org-drill-goto-drill-entry-heading' when no parent
heading carries the drill tag.
Coverage moved from 92.1% to 92.1%.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
I extended `tests/test-org-drill-route-rating-result.el' with four
`org-drill--pick-next-marker' cases (no resume → pop, resume with live
drill marker → keep current-item and clear resume-p, resume with nil
or non-drill current-item → fall through to fresh pop). I also
extended the resume regression file with the three happy-path branches
of `org-drill-resume': pending entries → resume, finished with pending
count → y-or-n-p offers a new session, finished with nothing → print
'finished'.
Coverage moved from 91.7% to 92.1%.
|
| |
|
|
|
|
|
|
|
|
|
| |
I added tests for `org-drill--route-rating-result' covering all four
return values: nil → 'quit (end-pos becomes :quit), 'edit → 'edit
(end-pos becomes a marker), 'skip → 'skip (current-item cleared),
quality 0 → 'next (pushes to again-entries) and quality 5 → 'next
(pushes to done-entries). Also a case where again-entries is non-empty
so the shuffle branch runs.
Coverage moved from 91.6% to 91.7%.
|
| |
|
|
|
|
|
|
|
|
| |
I added tests for `org-drill-leitner' with mocked `leitner-entry'
returning t (full loop completes, summary printed), 'quit (pcase quit
branch returns t), 'edit (pcase edit branch jumps to marker), and a
case where the boxed queue is short of `org-drill-maximum-items-per-session'
so `leitner-start-box' runs to top it up.
Coverage moved from 89.8% to 91.6% — the suite is now over 90%.
|
| |
|
|
|
|
|
|
|
|
|
| |
I added tests for the public `org-drill' command that mock
`org-drill-entries' so the orchestrator runs in batch: empty buffer →
'no pending' message, populated buffer → entry loop runs, cram=t flag
flips the session's cram-mode slot, resume-p skips entry collection.
I also extended the cloze + scheduling helpers file with sm2 and
simple8 coverage for `org-drill-hypothetical-next-review-date'.
Coverage moved from 89.0% to 89.8%.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
copy-to-buffer fallback
I added tests for `org-drill-replace-entry-text' with the multi-p flag
(list of replacements creates multiple overlays) and the simple
single-overlay case, the multicloze weight-validation error branches in
`-firstmore' and `-firstless', and the
`org-drill-copy-entry-to-other-buffer' recovery path that appends to
the end of DEST when the source's outline path doesn't exist there.
Coverage moved from 88.6% to 89.0%.
|
| |
|
|
|
|
|
|
|
|
|
| |
I added tests for `org-drill--maybe-prepend-leech-warning' (three
branches: not a leech, method not warn, leech with warn), the
presentation minibuffer-timer function (emits MM:SS prompt, cancels
after 10 calls, switches to '++:++' after an hour), the
`org-drill-cloze-length-matches-hidden-text-p' branch (display becomes
a dotted string), and the simple8 random-noise dispersal-factor branch.
Coverage moved from 87.8% to 88.6%.
|
| |
|
|
|
|
|
|
|
|
|
| |
I extended three existing test files with cases that hit branches the
suite was missing: `org-drill--show-end-message' with a live-marker
end-pos (jumps to the marker), `org-drill--restore-display' for the
variable-pitch-on / variable-pitch-off / text-scale paths, and
`org-drill-pop-next-pending-entry' for the young-mature and overdue
branches in the queue priority cond.
Coverage moved from 87.0% to 87.8%.
|
| |
|
|
|
|
|
|
|
|
| |
I added tests for `org-drill-leitner-start-box' (move N entries from
unboxed into box 1, respect the count arg, zero is a noop) and for
`org-drill--read-rating-key' (string input, arrow vector, wheel-event
vector, help key showing help block, tags key triggering set-tags,
typed-answer rendering in the prompt).
Coverage moved from 85.7% to 87.0%.
|
| |
|
|
|
|
|
|
|
| |
I added tests for `org-drill-merge-buffers' (yes/no confirmation, defaulting
dest to current buffer, full migrate pipeline), `org-drill-all-leitner-capture'
(populate and reverse boxed/unboxed queues), and the
`org-drill-leitner-vs-drill-entries' summary message.
Coverage moved from 82.4% to 85.7%.
|
| |
|
|
|
|
|
|
|
|
|
| |
scope=directory
I added unit tests for the display-state helpers (capture text scale,
variable-pitch, modeline; restore them on session exit), the directory
branch of `org-drill-current-scope', and `--migrate-from-source''s
three-branch cond (matching ID, no ID, ignore-new-items).
Coverage moved from 81.8% to 82.4%.
|
| |
|
|
|
|
|
|
|
| |
I added direct tests for `org-drill-map-leitner-capture' (route by
DRILL_LEITNER_BOX), the response-mode exit-kind handlers (quit/edit/skip/
tags/rtn), `org-drill-presentation-timer-cancel', and the sm2/simple8
branches of `org-drill-smart-reschedule''s algorithm dispatcher.
Coverage moved from 80.6% to 81.8%.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
I added 16 ERT tests for the helpers carved out of org-drill and
org-drill-merge-buffers in the recent refactor pass: prepare-fresh-session,
queues-empty-p, collect-entries, show-resume-hint, show-end-message,
build-dest-id-table, copy-scheduling-to-marker, and
strip-unmatched-dest-entries.
The clean-completion test for show-end-message binds
org-drill-save-buffers-after-drill-sessions-p to nil so the dispatcher
doesn't trip save-some-buffers' interactive prompt under batch ERT.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previously-bundled apple.jpg had unclear redistribution rights —
flagged in upstream issue #34 as a blocker for clean MELPA hosting
and downstream packaging.
Replaced with USDA 2026 Apple from the U.S. Department of Health
and Human Services' Dietary Guidelines for Americans 2025-2030,
sourced from Wikimedia Commons. Public domain under 17 USC §105
(work of the U.S. federal government), no attribution required.
Source PNG resized to 298x348 (matching the original's footprint)
and flattened to a white-background JPEG so the existing
spanish.org and robot/spanish-robot.org file: links keep working
without modification.
Documented the bundled-content provenance in assets/README.md
alongside the made-for-emacs-badge.svg.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two pragmatic changes after watching CI fail repeatedly:
CI matrix: dropped 28.2. Emacs 28 ships Org 9.5; Cask is supposed
to pull our declared org>=9.6 over the built-in but doesn't reliably
in this CI setup, and several test categories use APIs/behaviors
(cl-letf on signal-hook-function, eieio idioms, modern org-fold-*)
that don't quite work on 28. Practical floor is now Emacs 29
(ships Org 9.6 built-in). Matrix is 29.4 + snapshot.
Scheduler error tests: added skip-unless (>= emacs-major-version 30)
to the test-scheduler--should-cl-assert helper in both simple8 and
sm5 test files. ERT 29 installs an aggressive
signal-hook-function around the entire ert-deftest body that
intercepts every signal before any inner condition-case runs;
shadowing the hook locally doesn't help (verified across four
attempts). The eight cl-assert-precondition tests now run on
Emacs 30+ where ERT's hook leaves inner condition-case alone, and
skip on 29.x. All other tests still run on 29.4.
Locally green. Pushing to verify CI.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Attempt 4 at making the eight scheduler error tests pass on
Emacs 29.4 in CI.
The earlier approaches kept failing because ERT 29.4 installs
ert--should-signal-hook as signal-hook-function around the entire
ert-deftest body — not just inside should forms. That hook fires
on every signal before any inner condition-case can catch it,
which is why even a bare (condition-case ... (cl-assertion-failed
nil)) at the top of the test body didn't work.
The new helper rebinds signal-hook-function to nil inside its own
let-scope, so condition-case catches the cl-assertion-failed
signal normally. The ert-fail on the no-error path runs outside
that shadowing scope, so it still routes through ERT's failure
handling.
Locally green; pushing to test 29.4 in CI.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous fix wrapped condition-case in (should (eq 'caught ...)),
but ERT in Emacs 29.4 installs signal-hook-function around should
forms — that hook fires on every signal, intercepting them before
the inner condition-case can catch. CI on 29.4 still failed.
This iteration drops should entirely. Each test body becomes a
plain condition-case at the top level: run the form, and if it
returns normally, ert-fail. Catch cl-assertion-failed by name
rather than via the error parent — its parent-class registration
is inconsistent across Emacs versions, but the symbol-name match
through condition-case always works.
Locally green; let's see what 29.4 does with it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Four new top-level docs to round out the project's public-facing
materials:
- LICENSE: full GPL-3 text, mirroring the header notice in
org-drill.el (copied from emacs-wttrin's bundled license).
- CONTRIBUTING.md: bug-reporting guide, patch flow (fork → branch
→ tests → PR), development setup instructions wrapping the
existing make targets, and testing/style notes pointing at the
TDD discipline expected.
- CHANGELOG.md: keepachangelog format with an Unreleased section
documenting every change since the fork — bug fixes by issue
number, new infrastructure (CI, coverage, lint), structural
refactors, and removals.
- NEWS: shorter user-facing summary highlighting the bugs most
likely to have bitten existing users.
Closes the [#A] TODO 'Add CONTRIBUTING, CHANGELOG, NEWS, and a
standalone LICENSE file'.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The :type 'cl-assertion-failed' fix didn't help — Emacs 29.4 in CI
still marks the eight scheduler error tests as failures even though
the cl-assertion-failed signal clearly fires (visible in the
test-failure backtrace).
Whatever ERT's should-error is doing in 29.4, it isn't accepting
the signal as a pass. Replacing should-error with a manual
condition-case wrapped in should sidesteps the fragility — we
just verify SOMETHING was signalled, which is all the test ever
needed.
Extracted as a test-scheduler--should-cl-assert helper macro in
each file (the two test files don't share infrastructure right
now).
Locally green; expected to clear the 29.4 CI failure.
|
