This might actually OOS unless the command to set debug overlays is synchronised, I'll check.

Performance wise, on my computer I had measured (very imprecisely) about 0.2µs overhead for sending a message, which with enough entities could become 1ms or so. Won't be significant but will help anyhow.

In D454#18423, @wraitii wrote:

This might actually OOS unless the command to set debug overlays is synchronised, I'll check.+

Since rendering shouldn't change state (I sort of consider that not being const a small issue, but I'll not employ mutable to hack around this, and not storing mostly unchanged data seems wasteful) and since this is a debug option (and part of the simulation, so it shouldn't be changed in non-synchronized ways) I think it works, but please do check this. I'm putting this here because someone else might catch some issue.

Performance wise, on my computer I had measured (very imprecisely) about 0.2µs overhead for sending a message, which with enough entities could become 1ms or so. Won't be significant but will help anyhow.

This is a system component, so unlikely to result in lots of messages. But not doing pointless work is not doing pointless work, so there's that.

Build is green

Updating workspaces.
Build (release)...
Build (debug)...
Running release tests...
Running cxxtest tests (306 tests)..................................................................................................................................................................................................................................................................................................................OK!
Running debug tests...
Running cxxtest tests (306 tests)..................................................................................................................................................................................................................................................................................................................OK!

http://jw:8080/job/phabricator/1082/ for more details.

I'd need a RL test but this is not synchronised as far as I can tell, so it would OOS on a full hash I suppose (not familiar with how we handle OOS in MP).

This is a system component, so unlikely to result in lots of messages. But not doing pointless work is not doing pointless work, so there's that.

Ah right, brainfart :p

I tried some games in sp, mp with rejoin, with -serializationtest playing with overlay lines and didn't get oos.
That's not a proof and I have no real knowledge about those things.