Fix broken strings in localizations.

[jaydeetay]

The v2 rewrite salvaged v1's translated UI strings, and the migration refused 50 of them across 25 locales because the translation could not be trusted. Those are pre-existing defects in this repo, not v2 problems, so this marks them where they live for a human to fix or re-run through tm.

The bulk is the warm_welcome_slide3_* pair, merged badly at some point: the finished translation is followed by a repeated fragment of the English original, e.g. ja "...宇宙全体を探索できます！Manual Mode!" — visible English inside a Japanese string. 21 locales have that trailing-tail form, 23 have a longer untranslated run inside the text, 2 (de, fa) show a stray "=" splice joining translation to original, and 4 have mismatched inline markup.

Note tm validate does not catch most of these: the tag set is intact, so a complete ... beside half-translated prose passes. Error counts are unchanged by this commit (verified de/ja/fr before and after) — these are comments only, 100 insertions and no deletions.

Comments carry the SALVAGE-FLAGGED marker so they are greppable and so the generator can replace rather than duplicate them on a re-run. Generated by stardroid-v2's tools/salvage-translations/annotate_v1.py, which re-derives the list from the salvage tool's own verdicts rather than a hardcoded table.

Resources still merge (:app:mergeGmsDebugResources) and every strings.xml still parses.

Description

Type of Change

• Bug fix
• New feature (Sky Map team only)
• Translation (new or updated)
• Documentation
• Refactoring / code cleanup
• Dependency upgrade
• Other

Checklist

• I've read the contributing guidelines
• For major changes, I've emailed skymapdevs@gmail.com first
• I've run the unit tests with ./stardroid-v1/gradlew :app:test
• I've tested on a device/emulator if applicable
• If I have multiple commits, I've squashed them into one

No need to update CHANGELOG.md - we'll get Claude to do that when we cut a release.

Notes for Reviewers

[jaydeetay]

@claude please review

[claude]

Claude finished @jaydeetay's task in 2m 6s —— View job

---

Review of PR #950

• Gather context (diff, changed files)
• Understand the request (code review only)
• Review the diff for correctness
• Post review feedback

Key finding: the PR description no longer matches the diff

This PR has two commits, and they contradict each other:

1. 2cd4a75 "Flag half-translated strings with SALVAGE-FLAGGED comments" — adds the `` comments described in the PR body (comment-only, no rendered-string changes).
2. ec7f928 "Fixed broken translations."` — deletes every one of those comments and directly rewrites the 25 translated strings instead.

The result: grep -rl "SALVAGE-FLAGGED" stardroid-v1/app/src/main/res/ on the current branch head returns zero files. None of the flagging described in the title/body survives — the PR title ("Flag half-translated strings with SALVAGE-FLAGGED comments"), the description ("marks them where they live for a human to fix or re-run through tm"), and the accounting ("100 insertions and no deletions") all describe commit 1 only and no longer describe what this PR actually does. git diff origin/master...HEAD --stat shows 50 insertions / 121 deletions across the 25 files — real content changes, not comments.

Concretely (e.g. values-de/strings.xml, values-fr/strings.xml), commit 2 removes the English tail/stray =/dropped <b> and re-wraps <b> around what looks like the correct translated span (e.g. German <b>manuellen Modus</b>, French <b>Mode Manuel</b>). Spot-checking several locales (ar, de, fa, fr, ja, tr), the fixes look plausible and XML stays well-formed, but:

• This conflicts with AGENTS.md: "translations to other locales will be done after features are implemented by a separate pipeline." Directly hand/tool-editing 25 locales' translated prose is exactly what that convention says not to do in this kind of PR — it bypasses tm and there's no indication a native speaker reviewed the reworded text (as opposed to just moving markup).
• No before/after tm validate or error-count verification is mentioned for commit 2, unlike commit 1 (which explicitly checked de/ja/fr).
• The PR's own checklist item "If I have multiple commits, I've squashed them into one" is unchecked — and here it matters, since the two commits aren't incremental, the second reverses the first's approach entirely.

Recommendation: decide which behavior this PR is actually meant to ship — comment-only flagging (revert commit 2, keep title/body as-is) or direct fixes (drop commit 1's comments as dead weight, rewrite the title/body to match, and get native-speaker/tm verification for the reworded strings) — then squash to one commit whose message matches the diff.

Minor

• Where commit 2 touches wording beyond markup placement (e.g. tr Manual Mod → Manuel Mod, figure-8/sekiz phrasing across locales), that's a substantive translation correction, not just tag repositioning — worth calling out explicitly in the PR body if kept.

Not verified

I could not run :app:testDebugUnitTest or :app:mergeGmsDebugResources in this environment (network/tool restrictions on running gradle/python here), so the "resources still merge" and "every strings.xml still parses" claims are based on manual inspection of the diff only, not a build run.