* 🐛 Fix library updates reappear after file is reloaded
Summary
Migrate synced_at timestamps to a standalone file_library_sync table to ensure sync state is tracked for both direct and transitive libraries.
Problem
Transitive libraries (libraries imported by other libraries) are not stored as direct rows in file_library_rel. Because the system previously coupled synced_at directly to the file_library_rel schema, transitive libraries lacked a persistent location for their sync timestamps. This caused sync states to be lost or incorrectly reported for nested dependencies.
Changes
Schema Migration: Created file_library_sync and migrated existing synced_at values from file_library_rel.
Decoupling: Removed tight Foreign Key coupling to allow sync rows to exist independently of specific relationship records.
Persistent Writes: Added upsert-file-library-sync! helper. Updated all import, duplication, and RPC write paths (v1/v2/v3 importers, link-file-library) to ensure every write persists a sync row.
Unified Reads: Updated both direct and recursive/transitive library queries to fetch synced_at from the new table.
Testing: Added regression tests to verify that sync rows are correctly created/updated even when a transitive relation is absent in file_library_rel.
Impact
This fix ensures that the system accurately records and retrieves sync states for the entire library dependency tree, resolving the bug where nested libraries appeared out of sync.
* ✨ MR review
Caused when file is deleted in the middle of an exportation. The
current export process is not transactional, and on file deletion
several queries can start return not-found exception because of
concurrent file deletion.
With the changes on this PR we allow query deleted files internally
on the exportation process and make it resilent to possible
concurrent deletion.
Replace general usage of virtual threads with platform threads
and use virtual threads for lightweight procs such that websocket
connections. This decision is made mainly because virtual threads
does not appear on thread dumps in an easy way so debugging issues
becomes very difficult.
The threads requirement of penpot for serving http requests
is not very big so having so this decision does not really affects
the resource usage.
We don't migrate the libraries for avoid cascade migration; it is not ideal
but it reduces the total of the required memory needed for process a single
file migration that requires libraries to be loaded.
The patch was already existed but only applied to binfile-v3,
with this commit, the fix is properly applied to all binfile
formats and for duplicate file operation.
* 📎 Set proper name to relink-refs mechanism function
* 🐛 Fix incorrect id assignation on snapshot file resolution
* ♻️ Use uniform api for file retrieval on file snapshot code
The frontend-only features are now ignored from files and from teams
and they do not autoassigns automatically to team and file on file
creation or update operations.
Relevant changes:
- Add the ability to create migration in both directions, defaulting
to identity if not provided
- Move the version attribute to file table column for to make it more
accessible (previously it was on data blob)
- Reduce db update operations on file-update rpc method
Previously the file processing was implemented 3 times using similar
approaches bug each own with its own bugs. This PR unifies the
loging to a single implementation used by the 3 operations.
