NiLuJe

You'll get thrown a SQLITE_BUSY if Nickel is actively writing somewhere for some reason, so, providing both sides have a sane busy handler, it *should* behave ;). EDIT: And it's...

Sidebar about `SQLITE_BUSY`: I find the human-readable variant from [sqlite3_errstr/sqlite3_errmsg](https://github.com/sqlite/sqlite/blob/d63c76fb31a0e262fb12a93b171e95a4e30c1a2e/src/main.c#L1553-L1554) so incredibly confusing when compared to `SQLITE_LOCKED`, when the two actually mean *very* different things, that I kind of jumped...

Crappy workaround ideas ahead: Hmmm, something something SQL (inside a blocking shell script)? Along the lines of periodically checking if a specific TRIGGER has fired based on whatever the full_scan...

That also means that you can't actually block in a cmd_spawn/cmd_output action that isn't backgrounded (and if it is backgrounded, it means it can't be followed by a chain), right?...

IIRC, the issue with the `Activity` table is that it isn't "live": it's updated in batch according to arcane rules that no-one actually groks, I think?

I'll admit that it's been a good long while (as in, years) since I checked, but I remember abandoning that avenue quickly when I looked into it as a means...

Who's doing the reading, and who's doing the writing? If I get what you're trying to do, the frankenstein'ed NM is responsible for notifying about an event, so, it'd be...

@shermp: Hence the trickery with having the listener set it up first, opening it nonblock & rw ;). (See the stack overflow link in the FBInk comment for the full...

@shermp: Does the script actually return 0? Does NM actually get the right return code? @pgaskin: That made me double-check NM's doc re: chain: it reads like a chain can...

@shermp: https://github.com/shermp/Kobo-UNCaGED/blob/466f8ba92b54abd72df9c5565191c23eac62fecb/scripts/ku-lib.sh#L80 `oldifs` -> `$oldifs`?