Skip to content

fix: fixing a word's language no longer deletes it; docs say typed words reach Wiktionary (launch checks F1, F5) - #42

Merged
HackTuah merged 3 commits into
mainfrom
fix/relang-keeps-word
Oct 8, 2026
Merged

HackTuah merged 3 commits into
mainfrom
fix/relang-keeps-word

Conversation

@HackTuah

@HackTuah HackTuah commented Oct 8, 2026

Copy link
Copy Markdown
Member

Two fixes from the automated launch checks on v1.0.0-rc.4. Every main flow passed in Firefox and Chromium (20/20 each), using the release zips. These two are the findings worth fixing before the tag.

  • F1 (bug): "Another language" could delete the word. When the model answered with the same word in the same language ("Already in your list"), the new add landed on the old add's word, and retiring the old add then deleted it. The word went to Recently deleted while the popup said it was in the list. retireReplaced now keeps any word the new add landed on. The regression test in test/bg/local-mode.test.mjs re-adds a word as itself: before the fix the list was [], after it ["es:gato"]. The existing test (a real change of language removes the old word) still passes.
  • F5 (docs): typed words do reach Wiktionary. The docs site and README said nothing leaves your browser for word = meaning, but for a language with word stress Kotiko looks the word up on English Wiktionary for its pronunciation, as the privacy policy already says. Those pages now say so.
  • The security report now names this one code change after rc.4. It touches no security control, so the gate stays closed; the lead checked with check-security-gate.mjs --version 1.0.0.

Filed for 1.0.x instead:

  • F2: Recently deleted on an open dashboard is stale in local mode;
  • F3: focus is lost after the M menu;
  • F4: Enter in the popup's language search does nothing;
  • F6: the plain-HTTP warnings still mention the access key.

Checks: npm test 2082 pass; lint; site npm test 25/25 and npm run check; old-name; DCO.

🤖 Generated with Claude Code

Ayla Croft and others added 3 commits October 7, 2026 20:27
…e answer is the same word

Launch check F1 on rc.4: "Another language" queues a new add that
replaces the old one, and once it's saved the old add is undone. When the
model answered with the same word in the same language ("Already in your
list"), the new add landed on the old add's word, and undoing the old add
deleted it: the word left the list for Recently deleted while the popup
said it was already there. Retiring the old add now keeps any word the new
one landed on. A regression test re-adds a word as itself: before, the
list was empty; now the word stays. The existing test (a real change of
language removes the old word) still passes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Ayla Croft <dev@scriptkittyos.com>
Launch check F5: the docs site and the README said nothing leaves your
browser when you type 'word = meaning', but for a language with word
stress Kotiko looks up the word on English Wiktionary for its
pronunciation. The privacy policy already said so; now these pages do too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Ayla Croft <dev@scriptkittyos.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Ayla Croft <dev@scriptkittyos.com>
@HackTuah
HackTuah requested a review from minitru as a code owner October 8, 2026 00:27
@HackTuah
HackTuah merged commit 719f26c into main Oct 8, 2026
16 of 27 checks passed
@HackTuah
HackTuah deleted the fix/relang-keeps-word branch October 8, 2026 01:02
@HackTuah HackTuah mentioned this pull request Oct 8, 2026
9 of 11 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant