Translations 2/2: language choice in Settings, contexts, one-pass placeholders (#14, #33) #136

Merged

claude-bot merged 5 commits from i18n/part-2 into main 2026-10-30 15:36:52 +00:00

Conversation 3 Commits 5 Files changed 24 +511 -85

claude-bot commented 2026-10-30 14:49:47 +00:00

Collaborator

• Copy link
• Copy Markdown

What

You can now choose the app's language in Settings (the phone's language, English, or any installed catalogue, each named in its own language). The change shows at once with no restart, and the choice is kept. The gettext catalogue now keeps message contexts, so qsTr("…", "comment") is translated, and a string that isn't translated falls back to its source text, never to blank. tr_args fills %1–%9 in one pass, so a value that contains %2 is never filled in again. CI's e2e job now gets the catalogues along with the binary.

Closes #33. Refs #14 (it stays open for phone QC). Plural forms and languages beyond German are deferred to #135.

Blast radius

Tier: trunk
Trunk paths touched:

• core/src/worker*
• core/src/tasks*
• qt/main.cpp, qt/src/i18n.rs, qt/src/reading_model.rs
• qml/Main.qml
• tools/e2e/cases/language-switch.sh
• .forgejo/workflows/ci.yml

Testing tier: maximum for .forgejo/workflows/ci.yml (the e2e artifact). The rest is standard.
Gate / rollback: ui.language defaults to "" (the phone's language). On that path start-up is unchanged: the same init(), and no second catalogue. Undo: revert this merge.

Proof

• Local gates:

  • cargo make: fmt, clippy, the tests (utsb-core 291 passed), release, deny and audit all passed.
  • cargo make app-docker: exit 0.
  • tools/i18n/check-app.sh: the window title (a context string, "Study Bible", "window title") shows "Studienbibel" under LANGUAGE=de and "Study Bible" in the C locale.

• Fails without the change: with the .mo parser skipping context entries again, 2 of the i18n tests fail. Restored, all 9 pass. tr_args is a new API with no base version.

• CI:

  • Run 264 failed only in language-switch: the app listed 2 options, because the artifact held only the binary, without share/locale.
  • Since a65a67a the artifact has 2 files (binary and catalogue), as run 265's download shows.
  • Run 266 on ed7ddb4: green on changes, rust, audit, app, e2e (with language-switch) and click.

• Phone (Plinius), build i18n/part-2@a65a67a93 and then ed7ddb4:

  • Settings shows "Phone language", English and Deutsch.
  • Tapping Deutsch makes the section "Sprache" at once (log utsb: language heading Sprache).
  • After a restart the footer reads "‹ Zurück", so the choice is kept.
  • Phone language brings English back.
  • The first build centred the option labels under the Suru style. ed7ddb4 fixes that; both builds are shown in the shots below.

  
  
  

Resources

• CI mem-check: peak PSS 165244 kB (slow profile 168434 kB), against a budget of 176000 kB. Text was on screen after 315 ms, against a budget of 3000 ms.
• Phone, German chosen, after scrolling Prov 6 to 8: PSS 63113 kB, of which 42440 kB is anonymous. The budget is 80 MB (docs/07).
• Extra memory from a switch: one more parsed catalogue (tens of KB), only after the user switches. It is leaked once per language, so the leak is bounded by the languages installed.

#14 exit criteria

• Golden tests for 5+ Bibles over the KJV, MT and German versifications: tools/golden covers BSB, Darby, KJV, TR (KJV), OSHB (MT), and GerElb1871 and GerSch (German), with the v11n tables v11n-KJV, v11n-MT and v11n-German.
• Within budget on the Plinius: 63 MB against 80 MB, above.

Mockup

No mockup. The language choice is a Settings section of plain Controls 2 RadioButtons with a heading and a hint line, like the Navigation and Module repositories sections. It adds no page, sheet or dialog.

Review

Reviewer:

• Fresh-context review at high on e8eb225: 0 findings.
• Delta review at medium (360) on a65a67a, the CI artifact fix: 0 findings.
• ed7ddb4 is a one-line QML layout fix found on the phone. It comes after the reviews.

Review-Confidence: 100/100 (audit). Every trunk file was reviewed at high, and the CI-only change got a delta review. ed7ddb4 touches only a leaf page (SettingsPage.qml); it was checked on the phone, and CI is green. 0 findings are open.

Confidence by part

• tr_args and contexts: high. Unit tests, check-app, and fails without the change.
• Live switch: high. The e2e case and the phone.
• CI artifact: high. The e2e on ed7ddb4 is green.
• Radio label alignment: medium. Checked on the phone only; no e2e checks layout.

🤖 Generated with Claude Code

## What You can now choose the app's language in Settings (the phone's language, English, or any installed catalogue, each named in its own language). The change shows at once with no restart, and the choice is kept. The gettext catalogue now keeps message contexts, so `qsTr("…", "comment")` is translated, and a string that isn't translated falls back to its source text, never to blank. `tr_args` fills `%1`–`%9` in one pass, so a value that contains `%2` is never filled in again. CI's e2e job now gets the catalogues along with the binary. Closes #33. Refs #14 (it stays open for phone QC). Plural forms and languages beyond German are deferred to #135. ## Blast radius **Tier:** trunk **Trunk paths touched:** - `core/src/worker*` - `core/src/tasks*` - `qt/main.cpp`, `qt/src/i18n.rs`, `qt/src/reading_model.rs` - `qml/Main.qml` - `tools/e2e/cases/language-switch.sh` - `.forgejo/workflows/ci.yml` **Testing tier:** maximum for `.forgejo/workflows/ci.yml` (the e2e artifact). The rest is standard. **Gate / rollback:** `ui.language` defaults to "" (the phone's language). On that path start-up is unchanged: the same `init()`, and no second catalogue. Undo: revert this merge. ## Proof - **Local gates:** - `cargo make`: fmt, clippy, the tests (utsb-core 291 passed), release, deny and audit all passed. - `cargo make app-docker`: exit 0. - `tools/i18n/check-app.sh`: the window title (a context string, `"Study Bible", "window title"`) shows "Studienbibel" under `LANGUAGE=de` and "Study Bible" in the C locale. - **Fails without the change:** with the `.mo` parser skipping context entries again, 2 of the i18n tests fail. Restored, all 9 pass. `tr_args` is a new API with no base version. - **CI:** - Run 264 failed only in `language-switch`: the app listed 2 options, because the artifact held only the binary, without `share/locale`. - Since a65a67a the artifact has 2 files (binary and catalogue), as run 265's download shows. - Run 266 on ed7ddb4: green on changes, rust, audit, app, e2e (with `language-switch`) and click. - **Phone (Plinius), build `i18n/part-2@a65a67a93` and then ed7ddb4:** - Settings shows "Phone language", English and Deutsch. - Tapping Deutsch makes the section "Sprache" at once (log `utsb: language heading Sprache`). - After a restart the footer reads "‹ Zurück", so the choice is kept. - Phone language brings English back. - The first build centred the option labels under the Suru style. ed7ddb4 fixes that; both builds are shown in the shots below.    ## Resources - **CI mem-check:** peak PSS 165244 kB (slow profile 168434 kB), against a budget of 176000 kB. Text was on screen after 315 ms, against a budget of 3000 ms. - **Phone, German chosen, after scrolling Prov 6 to 8:** PSS 63113 kB, of which 42440 kB is anonymous. The budget is 80 MB (docs/07). - **Extra memory from a switch:** one more parsed catalogue (tens of KB), only after the user switches. It is leaked once per language, so the leak is bounded by the languages installed. ## #14 exit criteria - **Golden tests for 5+ Bibles over the KJV, MT and German versifications:** `tools/golden` covers BSB, Darby, KJV, TR (KJV), OSHB (MT), and GerElb1871 and GerSch (German), with the v11n tables `v11n-KJV`, `v11n-MT` and `v11n-German`. - **Within budget on the Plinius:** 63 MB against 80 MB, above. ## Mockup No mockup. The language choice is a Settings section of plain Controls 2 RadioButtons with a heading and a hint line, like the Navigation and Module repositories sections. It adds no page, sheet or dialog. ## Review **Reviewer:** - Fresh-context review at high on e8eb225: 0 findings. - Delta review at medium (360) on a65a67a, the CI artifact fix: 0 findings. - ed7ddb4 is a one-line QML layout fix found on the phone. It comes after the reviews. **Review-Confidence:** 100/100 (audit). Every trunk file was reviewed at high, and the CI-only change got a delta review. ed7ddb4 touches only a leaf page (`SettingsPage.qml`); it was checked on the phone, and CI is green. 0 findings are open. ## Confidence by part - **tr_args and contexts:** high. Unit tests, check-app, and fails without the change. - **Live switch:** high. The e2e case and the phone. - **CI artifact:** high. The e2e on ed7ddb4 is green. - **Radio label alignment:** medium. Checked on the phone only; no e2e checks layout. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude-bot added 3 commits 2026-10-30 14:49:47 +00:00

i18n: tr_args fills placeholders in one pass; the .mo parser keeps contexts (#33) ... e768c88b6b

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs: translations part 2 plan; the language choice is no longer the translators' work (#14) ... 91ff77aec7

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

i18n: choose the app's language in Settings, live; qsTr contexts; extraction of tr_args (#14, #33) ... e8eb225b41

Setting ui.language, Settings > Language, catalogue swap with QQmlEngine::retranslate,
msgctxt in the QTranslator and in xgettext, the window title as a context message,
check-app.sh and the language-switch e2e case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude-review reviewed 2026-10-30 14:51:55 +00:00

claude-review left a comment

• Copy link
• Copy Markdown

Fresh-context review. Level: high. Blast radius: trunk. Reviewed commit: e8eb225b41. Range: origin/main...HEAD.

Findings: none (no high, medium or low defect with a concrete failure scenario).

What was checked, in context: the Catalog context keys (\x04) and get_in against QML's empty-string disambiguation and the C++ null; select/load/CHOSEN/LOADED locking and the path guard on tags; the "" / "en" / unknown-tag fallbacks; tr_args in one pass against the old chained .replace (same results for %%, a value holding %N, a missing value, multi-byte text); every converted call site in tasks, worker and session (same placeholders, same order); the ui.language setting through SetSetting, set_setting (render.apply ignores the key), settings() at start and ReadingModel::initialize (a single ReadingModel instance, so the default "" cannot reset a chosen language); choose() on the GUI thread and the retranslate callback registered before engine.load; the SettingsPage options order against the e2e case's 6 coordinates; the xgettext keyword specs and the perl extraction of tr_args against po/ and the pot; check-app.sh against the msgctxt entry. cargo test -p utsb-core --lib i18n passes locally (9 tests). CI on this commit was still pending at review time.

Observations, not defects: strings the worker made before a language change (notices, task row texts already sent to QML) keep the old language until the next event; the code comment and docs say so. Plurals stay with #135.

Files reviewed: core/src/i18n.rs, core/src/settings.rs, core/src/starter.rs, core/src/tasks.rs, core/src/tasks/exec.rs, core/src/tasks/progress.rs, core/src/worker.rs, core/src/worker/session.rs, core/src/worker/session/jump.rs, core/src/worker/session/switch.rs, qt/main.cpp, qt/src/i18n.rs, qt/src/reading_model.rs, qml/Main.qml, qml/pages/SettingsPage.qml, qml/pages/ReadingPane.qml (usage), tools/e2e/cases/language-switch.sh, tools/e2e/lib.sh (helpers), tools/i18n/update.sh, tools/i18n/check-app.sh, po/de.po, po/utstudybible.schaefera.pot, docs/09-quality-gates.md, docs/superpowers/plans/2026-10-21-gettext.md, docs/superpowers/plans/2026-10-30-gettext-2.md, CLAUDE.md

Fresh-context review. Level: high. Blast radius: trunk. Reviewed commit: e8eb225b4100e1f65a0c40e1563fb8cde2663baf. Range: origin/main...HEAD. Findings: none (no high, medium or low defect with a concrete failure scenario). What was checked, in context: the Catalog context keys (\x04) and get_in against QML's empty-string disambiguation and the C++ null; select/load/CHOSEN/LOADED locking and the path guard on tags; the "" / "en" / unknown-tag fallbacks; tr_args in one pass against the old chained .replace (same results for %%, a value holding %N, a missing value, multi-byte text); every converted call site in tasks, worker and session (same placeholders, same order); the ui.language setting through SetSetting, set_setting (render.apply ignores the key), settings() at start and ReadingModel::initialize (a single ReadingModel instance, so the default "" cannot reset a chosen language); choose() on the GUI thread and the retranslate callback registered before engine.load; the SettingsPage options order against the e2e case's 6 coordinates; the xgettext keyword specs and the perl extraction of tr_args against po/ and the pot; check-app.sh against the msgctxt entry. `cargo test -p utsb-core --lib i18n` passes locally (9 tests). CI on this commit was still pending at review time. Observations, not defects: strings the worker made before a language change (notices, task row texts already sent to QML) keep the old language until the next event; the code comment and docs say so. Plurals stay with #135. Files reviewed: core/src/i18n.rs, core/src/settings.rs, core/src/starter.rs, core/src/tasks.rs, core/src/tasks/exec.rs, core/src/tasks/progress.rs, core/src/worker.rs, core/src/worker/session.rs, core/src/worker/session/jump.rs, core/src/worker/session/switch.rs, qt/main.cpp, qt/src/i18n.rs, qt/src/reading_model.rs, qml/Main.qml, qml/pages/SettingsPage.qml, qml/pages/ReadingPane.qml (usage), tools/e2e/cases/language-switch.sh, tools/e2e/lib.sh (helpers), tools/i18n/update.sh, tools/i18n/check-app.sh, po/de.po, po/utstudybible.schaefera.pot, docs/09-quality-gates.md, docs/superpowers/plans/2026-10-21-gettext.md, docs/superpowers/plans/2026-10-30-gettext-2.md, CLAUDE.md

schaefera added 1 commit 2026-10-30 15:09:23 +00:00

ci: the e2e job gets the app's catalogues with its binary ... a65a67a93d

The language-switch case lists the installed languages from share/locale
next to the binary; the artifact held only the binary, so CI found no
German (run 264).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude-review reviewed 2026-10-30 15:10:01 +00:00

claude-review left a comment

• Copy link
• Copy Markdown

Delta review (medium)

Commit: a65a67a93d
Range: e8eb225..a65a67a
Files: .forgejo/workflows/ci.yml

Findings: none.

Checked: upload-artifact@v3 with two paths (build-ci/utstudybible, build-ci/share/locale) roots the artifact at their least common ancestor, build-ci, so the download into app-amd64 gives app-amd64/utstudybible and app-amd64/share/locale//LC_MESSAGES/*.mo. CMakeLists.txt puts the .mo files in ${CMAKE_BINARY_DIR}/share/locale via the ALL translations target, which the binary depends on. core/src/i18n.rs locale_dir looks at exe.parent()/share/locale, which matches that layout. Nothing else uses app-amd64: only the e2e download step and UTSB_APP=$PWD/app-amd64/utstudybible, and click does not consume it. tools/ci/e2e.sh only chmods the binary, so the artifact's old shape is not depended on.

Delta review (medium) Commit: a65a67a93d2ea290aaa04c7fd6b19c3ef61ff324 Range: e8eb225..a65a67a Files: .forgejo/workflows/ci.yml Findings: none. Checked: upload-artifact@v3 with two paths (build-ci/utstudybible, build-ci/share/locale) roots the artifact at their least common ancestor, build-ci, so the download into app-amd64 gives app-amd64/utstudybible and app-amd64/share/locale/<lang>/LC_MESSAGES/*.mo. CMakeLists.txt puts the .mo files in ${CMAKE_BINARY_DIR}/share/locale via the ALL `translations` target, which the binary depends on. core/src/i18n.rs locale_dir looks at exe.parent()/share/locale, which matches that layout. Nothing else uses app-amd64: only the e2e download step and UTSB_APP=$PWD/app-amd64/utstudybible, and click does not consume it. tools/ci/e2e.sh only chmods the binary, so the artifact's old shape is not depended on.

schaefera added 1 commit 2026-10-30 15:16:31 +00:00

ui: the language options keep their labels beside the circle ... ed7ddb4aee

On the phone the Suru style centres a full-width RadioButton's label.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude-bot changed title from WIP: Translations 2/2: language choice in Settings, contexts, one-pass placeholders (#14, #33) to Translations 2/2: language choice in Settings, contexts, one-pass placeholders (#14, #33) 2026-10-30 15:36:01 +00:00

claude-review approved these changes 2026-10-30 15:36:34 +00:00

claude-review left a comment

• Copy link
• Copy Markdown

Merge gate on ed7ddb4. Reviews 359 (high, e8eb225) and 360 (medium delta, a65a67a, ci.yml) have 0 findings, so nothing is left to fix or refute. The only commit after them, ed7ddb4, is a one-line QML change in qml/pages/SettingsPage.qml (a leaf page): it drops Layout.fillWidth from the language RadioButtons, with a comment. It does not touch logic. CI on head is green on all six jobs of run 266: changes, rust, audit, app, e2e and click. The PR body's Proof has no unrun elevated, maximum or AI-behavior gate. The maximum-tier ci.yml change is backed by the failing run 264 and the green run 266, which includes the language-switch e2e case. The Review-Confidence of 100 comes from the audit. Approved.

Merge gate on ed7ddb4. Reviews 359 (high, e8eb225) and 360 (medium delta, a65a67a, ci.yml) have 0 findings, so nothing is left to fix or refute. The only commit after them, ed7ddb4, is a one-line QML change in qml/pages/SettingsPage.qml (a leaf page): it drops Layout.fillWidth from the language RadioButtons, with a comment. It does not touch logic. CI on head is green on all six jobs of run 266: changes, rust, audit, app, e2e and click. The PR body's Proof has no unrun elevated, maximum or AI-behavior gate. The maximum-tier ci.yml change is backed by the failing run 264 and the green run 266, which includes the language-switch e2e case. The Review-Confidence of 100 comes from the audit. Approved.

claude-bot merged commit c0acc250c9 into main 2026-10-30 15:36:52 +00:00

claude-bot referenced this pull request from a commit 2026-10-30 15:36:52 +00:00

Merge pull request 'Translations 2/2: language choice in Settings, contexts, one-pass placeholders (#14, #33)' (#136) from i18n/part-2 into main

claude-bot referenced this pull request 2026-10-30 15:37:38 +00:00

Translations (gettext) and CI resource check #14

Sign in to join this conversation.

Reviewers

No reviewers

claude-review

Labels

Clear labels   

priority

Could

MoSCoW: nice to have, done when time allows

  

priority

Must

MoSCoW: the release isn't viable without it

  

priority

Should

MoSCoW: important, but the release works without it

  

priority

Won't

MoSCoW: not this release (kept for later)

  

type

Bug

No labels

priority

Could

priority

Must

priority

Should

priority

Won't

type

Bug

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

3 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

schaefera/utStudyBIble!136

No description provided.