If the user explicitly asked for 1 thread don't add an interactive one.

[gbaraldi]

This basically makes the assumption that if the user asked for 1 thread they really want their program to not have threads

[IanButterworth]

As this introduces a complexity to the setting, I think this need the docs to be updated to understand how clear this would be i.e. both -t and JULIA_NUM_THREADS behavior

[IanButterworth]

Makes sense to me, just needs a test alongside the ones that check threadpool sizes.

[vtjnash]

This sort of thing seems very unintuitive. Why are we getting hidden unexplained threads at all when the user didn't ask for them?

[IanButterworth]

Because the default number of interactive threads changed. See motivation in #57087

From discussion in the multithreading call there was general agreement that while this is complication of a simple rule, -t1 does seem like a case worth special casing:

-t2 : most likely you would benefit from separating main task (& io loop), and tasks on worker threads.
-tauto: same as above
-t1: you probably want specifically one thread because otherwise you'd just leave the default -t1,1

[IanButterworth]

Tests should pass once #58619 lands

[topolarity]

@gbaraldi Can you assist with backporting this?

[gbaraldi]

Yeah, I imagined this would apply quite cleanly