Manual refresh keeps stale file metadata after concurrent folder refresh inserts duplicate cache rows (35.0.1)

Aperta
#17,836 3 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
35/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
android, kotlin
Ambito
databases, mobile

Direzione di ricerca

Read FileContentProvider.upsertSingleFile(), RefreshFolderOperation.prefillLocalFilesMap(), and OCFileExtensions.filterFilenames() at stable-35.0.1. Trace how concurrent folder refreshes insert and select rows for the same account/path, then inspect the existing database and folder refresh tests. Done means refreshes show current metadata, preserve local download and sync state, and leave one cache row per account/path, including when legacy duplicates exist.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

0. Needs triage bug
⚠️ Before posting ⚠️
  • This is a bug, not a question or an enhancement.
  • I've searched for similar issues and didn't find a duplicate.
  • I've written a clear and descriptive title for this issue, not just "Bug" or "Crash".
  • I agree to follow Nextcloud's Code of Conduct.
Steps to reproduce
  1. On Android 35.0.1, browse a previously uncached directory while MetadataWorker is fetching that directory in the background. This timing-dependent race was captured twice on a real phone.
  2. The foreground RefreshFolderOperation and MetadataWorker both prepare inserts for the same account/path before either batch commits. Check the local public provider: two filelist rows now have the same account, path and remote_id.
  3. On the server, overwrite a PDF at the same path with a larger, newer PDF; keep its server fileid unchanged. Also rename a sibling directory and add a new file.
  4. Pull down to refresh the parent directory, wait for completion, navigate away/back, then refresh again.
  5. Compare the visible size/date and local provider rows with authenticated DAV PROPFIND and GET results. No account removal, app-data clearing, or reinstall is necessary for reproduction.
Expected behaviour

One completed manual refresh should show the current directory names and file metadata. A same-path replacement should show the current size/date. Cache identity should be unique per account/path, and concurrent metadata writes must preserve local download and folder-sync state.

Actual behaviour

The list continues to show the old PDF size/date through repeated manual refreshes. In the second controlled reproduction, cache row 1455 remained at 1584 bytes and the old modification time, while row 1457 for the SAME account/path/remote_id was updated to 2649 bytes and the new modification time. Both media_path values were NULL: this was not a stale downloaded PDF. The server fileid remained 33307.

A renamed old directory survived the first refresh because only one duplicate was scheduled for removal; another refresh removed it. A fresh fetch through Recent files updated the first file row and the list then showed the new size. Opening/downloading the PDF separately retrieved the correct newer bytes.

The same stale-list symptom also occurred when publishing by moving the old file into history and creating a new file at the original path, so a changed server fileid is not required.

Android version

16

Device brand and model

realme Neo7 (RMX5060)

Stock or custom OS?

Stock

Nextcloud android app version

35.0.1, versionCode 350000190, Google Play

Nextcloud server version

34.0.4.1

Using a reverse proxy?

Yes

Android logs

Scoped logcat excerpt, directory path anonymized:

10-18 21:28:57.933 22888 30602 D FileDataStorageManager: Saving folder /RefreshRace/ with 2 children and 0 files to remove
10-18 21:28:57.937 22888 30602 D FileDataStorageManager: Sending 3 operations to FileContentProvider
10-18 21:28:57.937 22888  3242 D FileDataStorageManager: Saving folder /RefreshRace/ with 2 children and 0 files to remove
10-18 21:28:57.940 22888  3242 D FileDataStorageManager: Sending 3 operations to FileContentProvider

Thread 30602 was MetadataWorker. Thread 3242 was foreground folder refresh. Public provider snapshots confirmed the duplicate rows after these batches. Full logs contain private filenames and are retained locally rather than posted wholesale.

Server error logs
Authenticated PROPFIND returned 207 with the current size, modification time and ETag. Authenticated GET returned 200 with the expected current PDF bytes. Local loopback and the public reverse-proxy route matched. Full server application logs are not included; the evidence here is the authenticated DAV responses and scoped access-log correlation.
Additional information

Source inspection at stable-35.0.1 (e6a4ade5047a90992a70652d0d5df658fd9a7c82):

  • FileContentProvider.upsertSingleFile() calls INSERT with CONFLICT_IGNORE first. FileEntity only has the autogenerated _id primary key; there is no unique constraint on account/path, so this insert does not conflict. Serialized applyBatch transactions alone cannot prevent duplicate rows when both batches were prepared before the first commit.
  • RefreshFolderOperation.prefillLocalFilesMap() collapses duplicates by remote path and refreshes one row.
  • OCFileExtensions.filterFilenames() uses distinctBy(fileName), so the first stale row can hide the updated row.

Candidate correction: atomic account/path lookup-and-write, preservation of existing local download/sync columns on a raced insert, newest-fetched metadata selection for legacy duplicates, and removal of every duplicate for a directory path absent from the refreshed listing. No destructive database migration or account reset is proposed.

Validation of the candidate build is still pending; the observations above are from the original signed Google Play client. The source analysis and candidate patch were prepared with OpenAI Codex assistance under the account owner's explicit direction. No human code review or DCO certification is claimed.

Lingua principale
Kotlin
Stelle
5.7k
Fork
2.1k
Merge medio
1g 22h
PR unite (30g)
135

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di nextcloud/android

Tutte le issue di nextcloud/android

Issue simili

Altre issue su Kotlin

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.