NiLuJe

Also, I imagine using `exec` in the `cmd_spawn` would get rid of one of those ;).

IIRC, the only way the database will be (write) locked on a WAL db is if someone (i.e., Nickel) is actively *writing* to it (possibly just committing, actually?). IIRC, there's...

https://sqlite.org/rescode.html#busy mentions using `BEGIN IMMEDIATE` to at least avoid getting a write lock in the *middle* of a transaction, which sounds neat/useful ;).

Forcing WAL checkpointing (https://www.sqlite.org/c3ref/wal_checkpoint_v2.html) might also help?

sickel *should* be murdering a hung or busy-looping Nickel, AFAIK? (I don't recall the exact details of how the sickel -> nickel ping works :s). What did things look like...

Re: sickel Possibly a dbus ping? In any case, if I SIGSTOP nickel, sickel will reboot after ~1 minute: ``` 01:13:45.975967 [2c3b2f44] pselect6(7, [3 6], [], [], {tv_sec=9, tv_nsec=980000000}, {NULL,...

Also, random comment about something that came up at the beginning of this thread: I imagine the koreader images could effectively serve as a "dockerized" variant of every koxtoolchain TC...

The KOReader images are definitely taking the "one image per TC" approach ^^. c.f., https://hub.docker.com/u/koreader

IIRC, Kobo doesn't ship the SQLite CLI tool. (It certainly didn't use to, as I deemed it safe to ship a symlink to mine in the Python packages. [As for...

Short of the (optional) ICU dependency, it can be built without readline (IIRC), and there's no longer any third-party deps at that point ;). (Short of a build dep: tcl...