Reasons for SQLite error 9: interrupted

[tgrapperon]

I'm using a few long-lived ValueObservation queries. From time to time, they fail with SQLite error 9: interrupted. Is it what I should expect if the observation is cancelled while it's currently fetching new results, or could there be other reasons for that error to be thrown? (for example the app being suspended or something in that vein)
I'm using GRDB 7 BTW, and the error is logged from the handleEvents(didFail:) method.

[groue]

Hello @tgrapperon,

SQLITE_INTERRUPT (9) is thrown when sqlite3_interrupt() is called.

GRDB calls this function when:

• The interrupt() method is called.
• The database gets "suspended", which happens if you set the observesSuspensionNotifications Configuration flag, and the app posts a suspendNotification Notification (see How to limit the 0xDEAD10CC exception).
• A cancelled async database access.

ValueObservation is not supposed to be affected by the third case (cancelled database accesses). I'd look if your app can trigger any of the two previous cases.

[tgrapperon]

Yes, the app uses in its vast majority async methods. Almost all reads and writes are async. Same for ValueObservation where I'm iterating over its values async sequence. It's also a very busy app if that matters. Dozens of concurrent accesses to the database in some cases.
I unfortunately don't have a reproduction myself. I have reports of updates stopping and logs of that error being thrown, but I didn't manage to trigger it myself in a debugging setup. It's also not systematic, but not completely rare either.
I plan to implement a sentinel observation with more logging to check if I can give more context to the problem.

[groue]

I wrote a little app that creates a lot of tasks that write, observe, and get cancelled, and I could reproduce the error. Thank you @tgrapperon, you just reported a bug (and a nasty one) 👍

[groue]

The fix: #1747

[tgrapperon]

Thank you so much from working on so little information!

[groue]

Thank you so much from working on so little information!

You're welcome! Thanks for reporting this issue. Observation ought to be as reliable as possible, and I'm somewhat embarrassed. So thank you again.

I believe the bug is now fixed in v7.4.0!