fix(sync): explain moves blocked by Trash

Keep blocked moves queued so they can land after a restore. Retain
simulation logs for intermittent conflicts and pick up the SDK fix for
uploads from machines with fast clocks.
This commit is contained in:
nvrl committed 2026-11-19 05:01:52 +01:00
1 parent 7d9d5850a4
commit c57d98799a
13 files changed
+222 -31

No files matched your search

+6 -1
View File
@@ -988,11 +988,16 @@ timing.
**Where:** not located; the drain's conflict check in `crates/pdfs-fuse/src/drain.rs` is the
first suspect
**Repro:** Only the CI run above. The last steps created `x/z/y/a.txt` (37 bytes), wrote 42
**Repro:** The first CI failure created `x/z/y/a.txt` (37 bytes), wrote 42
bytes into it at offset 20, and brought every link up. Drive then held the file at its first
37 bytes, and the full content was in `x/z/y/a (sync-conflict …).txt`. No other client wrote
to it.
On 2026-11-19, seed 1 of `one-client-echoes` showed the same symptom with delayed,
duplicated, and reordered events. A two-byte overwrite of `x/d` appeared only in a
conflict copy. Local replays passed. The simulation now retains its operation trace
and daemon logs on failure. [CI run](https://github.com/narrrl/proton-drive-linux/actions/runs/44427290348)
**Cause.** Not known. Nothing was lost, since the conflict copy holds the bytes, but the file
conflicted with its own create. B183 and B184 found the same symptom after a lost create
reply.
+13 -3
View File
@@ -13,13 +13,22 @@ scratch (user data in `staging/` and `recovery/` is never touched by this).
## [3.8.1] — 2026-11-19
### Added
Schema: **45**. No database migration in this release.
- **Branding** icons updated, screenshots updated
### Changed
- Refreshed the application icon and project artwork, and updated the documentation screenshots.
### Fixed
- **Cache Eviction** fixed an issue where it didn't scan past pinned files and now also reapplies the limit at startup.
- The cache skips pinned files when looking for space to reclaim and reapplies the configured
limit when the service starts. Pinned files and recovery data are preserved.
- Trashing a folder waits for queued moves out of it, including moves from nested subfolders.
If a move is blocked by an item or folder already in Trash, Sync and `pdfs sync issues`
identify what to restore before retrying.
- The system tray displays a warning icon when sync needs attention or the session has expired.
- Updated the Proton SDK to 0.12.1 so a fast local clock no longer causes Proton to reject
uploads of new files and photos.
## [3.8.0] — 2026-11-14
@@ -2363,6 +2372,7 @@ First stable release: FUSE files-on-demand mount, sync daemon under `proton-driv
- The outstanding FUSE defects tracked in `docs/BUGS.md`, plus a truncate defect, validated
by a new POSIX compliance suite for the filesystem.
[3.8.1]: https://github.com/narrrl/proton-drive-linux/releases/tag/v3.8.1
[3.5.0]: https://github.com/narrrl/proton-drive-linux/releases/tag/v3.5.0
[3.4.1]: https://github.com/narrrl/proton-drive-linux/releases/tag/v3.4.1
[3.4.0]: https://github.com/narrrl/proton-drive-linux/releases/tag/v3.4.0
+2 -1
View File
@@ -75,7 +75,8 @@ PDFS_SIM_SEED=7 RUST_LOG=pdfs_fuse=debug cargo test -p pdfs-fuse --lib \
| `PDFS_SIM_MEASURE` | Run `a_thousand_files_drain_as_fast_on_wifi_as_on_lan`, which times 1,000 files in 100 folders on a LAN and a Wi-Fi profile and prints the times |
A failure is a seed that can be replayed, but the daemon uses real threads, so a race may need
several replays. In CI, the `simulation` job uploads the stacks of a hung run as an artifact.
several replays. In CI, the `simulation` job uploads the failed seed's steps and daemon logs,
along with thread stacks when the watchdog catches a hang.
## Acceptance suite