Fix bug causing worker-initiated messages to be broadcast to every map instance #3133
Conversation
|
@jfirebaugh This should fix it. Confirmed it with a quick check using jsfiddle example linked in the issue. |
| send: function (type, data, callback, buffers) { | ||
| this.actor.send(type, data, callback, buffers, mapId); | ||
| }.bind(this) | ||
| }; |
There was a problem hiding this comment.
It feels like we're fighting our own architecture with this fake Actor. Can you think of a way to refactor & simplify? Worker and Actor are so tightly coupled it might make sense to merge them.
There was a problem hiding this comment.
I think merging Worker and Actor is probably beyond the scope of this bug fix, but I certainly agree your broader point over in #3034.
I was trying to keep the scope of this change as limited as possible, but maybe one way to simplify this a bit would be to replace WorkerSource's (and WorkerTile#parse's) actor: Actor parameter with just a send: (type, data, callback)=>void parameter.
|
Another issue with #2952 is that it can leak WebWorker Can you add a fix for that to this PR as well? |
|
Another issue: the blob URLs that webworkify creates aren't revoked; if you repeatedly create and remove maps, so that the worker pool is drained and refilled, they leak. |
I think this predates 2952, but yep I can add a fix for this. |
Ah, actually it probably doesn't predate 2952, since it's probably a pretty safe assumption that, before the worker pool, the |
anandthakker
force-pushed
the
fix-worker-messages
branch
from
4c2ab4b
to
cb2d45d
Compare
|
@jfirebaugh included fixes to dispose of event listeners and revoke object URLs. @lucaswoj also tried to clarify the code in |
lucaswoj
force-pushed
the
fix-worker-messages
branch
from
cb2d45d
to
80b2925
Compare
|
This is ready to 🚢. We need to fix this bug before the next release. I'm also keen on simplifying our worker architecture in a follow-up. |
Closes #3086