Daniel Markstedt

There’s a tricky bug right now where when you copy multiple files in one batch into a subdirectory, only the first file gets written into the database. This does not...

This practically works now. The big question to figure out right now is how to make the database writable by any AFP user while avoiding 666 permissions on the sqlite...

Pulling this back into draft again. The current solution works great for a single threaded scenario, when a single afpd process accesses the sqlite database. However, when multiple afpd forks...

The failing command is an INSERT (the _cnid_put_stmt_ prepared statement). This is a sample log if you're curious. It is curious to note that the logs suggest that .DS_Store doesn't...

@NJRoadfan I've refactored the code to use BEGIN + COMMIT / ROLLBACK consistently for destructive transactions, rather than relying on SQLite's automatic transaction handling. This doesn't resolve the database lock...

@NJRoadfan Read/write access by concurrent users works now. The key change was to explicitly call sqlite3_reset() on the prepared statement in the cleanup step of every CNID function call that...

~~There's a bug where moving a file from one folder to another folder leads to an undefined error in the logs. This doesn't have any visible impact on the AFP...

~~The spectest suite runs immensely better now, but we still have 9 failing tests out of the set that passes with the mysql backend.~~ These were fixed by improving error...

The full spectest is now running against the sqlite backend in CI jobs, and all tests are passing. Even those that are consistently failing with the mysql backend. ~~The final...

@NJRoadfan Would it make sense to merge this now, and roll forward with any fixes for `ad` compatibility that we discover later?