docs: align public documentation with current fork state

[antoinevalentinHA]

Summary

Documentation only. Three files, no code, no test, no tag, no release. The aim is that the public documentation describe what master actually does today — nothing more is claimed by it or by this PR.

README.md

Up to date through C6 and silent after it. What was corrected, and why:

Was	On master	Now
"164 Java tests, 172 with FAT32, 23 JavaScript"	282 and 57; the 172 is recorded in TESTING.md as last measured before the C6d-5 additions	282 and 57, with the FAT32 figure left to TESTING.md, which is the authority
"One documented session"	FIELD-VALIDATION.md records three, on two devices	three sessions, two devices
"retrying that upload has no defined semantics — it fails on the clear-text files"	the upload creates its folder exclusively and refuses, with a documented message	the refusal is described; the remaining gap — no supported way to resolve the state — is kept
"nothing detects or reports partial states"	getUnreferencedPackContentFolders() exists, with no caller outside tests	the listing exists, nothing calls it, nothing surfaces it, nothing acts on it

deletePack removing the index entry before the content is unchanged in the code and unchanged in the text: it was already exact.

Two behaviours were missing entirely and are now described: a pre-existing .content folder is refused rather than written into or emptied, and the library and conversion work that followed C6.

README_fr.md

It was the upstream README in full — no fork preamble, no statement of what the fork changes, no limitations, no field results. The same six sections as README.md are now present in French, written as French rather than translated word for word. The two files no longer diverge on installation, provenance, features, limitations, links, or release status.

Build instructions

Neither file named this repository. README.md told the reader that building from source was the way to use the fork and pointed at For developers, which clones marian-m12l/studio — so following its own instructions produced upstream's tree without any of the changes described above. Both files now give the clone URL of this repository, and the upstream URL is annotated where it appears rather than removed.

Tag and release status

There is no tag and no release for this fork: git ls-remote --tags origin returns nothing and the releases API returns zero. Both files say so. Both release badges are labelled as upstream's and carry a note; both download links are labelled as upstream builds. No link points at a release that does not exist, and no version number, badge or "stable release" wording was invented.

Conversion reuse

The upstream text in both files says the most recent file is transferred. That is no longer what happens — chooseDropAction compares no timestamps. The current behaviour is now documented at user level, in both languages: recency does not decide; when the library holds both something the device can read and something that could be converted into it, STUdio asks, unless it can show they correspond; new conversions record what they were made from and are re-read and compared before reuse; conversions made before that existed are never assumed to match.

No internal detail is exposed — no ledger format, no digests, no class names, no verdict names.

FIELD-VALIDATION.md

Six occurrences of third-party work titles, in two operation tables, replaced by neutral designations (personal pack 1–4, multi-track custom pack). They carried no technical information: the operation, device, UUID suffix, index count and outcome are in the other columns.

Everything else is untouched, and was verified by extracting the facts from both versions and comparing them: the six index-count transitions, both serial numbers, the firmware revision, the build's test count, the final device states, and the UUID suffixes. No result changed, no failure removed, no number, date, device or firmware altered. Two sentences were added where neutralising would otherwise have lost something — that two rows are the same pack sent to two devices, which the reuse observation depends on, and that these were personal packs holding third-party content rather than synthetic fixtures.

Not in scope

TESTING.md was audited and left alone: its counts are correct and its FAT32 figure is already dated honestly.

Three test files still contain the same titles as fixture strings — packs.test.js, DatabaseMetadataServiceResourceTest.java, ConversionProvenanceStoreTest.java. They are deliberately not touched here, since changing them is a code change. They are to be handled in a separate, bounded lot. This PR therefore makes the documentation free of those titles, not the repository.

Nothing here claims the fork is release-ready, and nothing repositions it relative to upstream. Attribution to marian-m12l/studio is unchanged.

🤖 Generated with Claude Code