Where are you starting? What can you see? We at Mozilla have seen some slow cycle collection times every now and then when using Element.
What do you click? N/A
More steps… This happens randomly and takes often some time to reproduce
Outcome
What did you expect? No leak
What happened instead? memory usage goes up and GC/CC times too
Copy-pasting my comment from elsewhere:
I may have gotten some useful information about the Element leak. Happened to see that locally. At least in this case lots of the memory stays around because there is some function n which through enclosing_environment and other stuff keeps openEvent alive, and that is a pointer to a FocusEvent. (openEvent is still marked). Now, FocusEvent isn't skippable, and its mRelatedTarget and mOriginalRelatedTarget point to an orphan html:li element. CC doesn't know about that FocusEvent, and thus html:li have two unknown refs. html:li does have a preserved wrapper and of course that then keeps lots of stuff around (because of react stuff). So, a leak on the site. We should be able to decrease CC time quite a bit, but the leak is still there.
So seems like there is this "openEvent" somewhere through which lots of memory stays around. Or something is keeping openEvent around. The path is 'function n->Call->LexicalEnvironment->a->current->openEvent->FocusEvent'
My guess something somewhere is forgetting to clear openEvent, or that current.
Steps to reproduce
Outcome
What did you expect? No leak
What happened instead? memory usage goes up and GC/CC times too
Copy-pasting my comment from elsewhere: I may have gotten some useful information about the Element leak. Happened to see that locally. At least in this case lots of the memory stays around because there is some function n which through enclosing_environment and other stuff keeps openEvent alive, and that is a pointer to a FocusEvent. (openEvent is still marked). Now, FocusEvent isn't skippable, and its mRelatedTarget and mOriginalRelatedTarget point to an orphan html:li element. CC doesn't know about that FocusEvent, and thus html:li have two unknown refs. html:li does have a preserved wrapper and of course that then keeps lots of stuff around (because of react stuff). So, a leak on the site. We should be able to decrease CC time quite a bit, but the leak is still there.
So seems like there is this "openEvent" somewhere through which lots of memory stays around. Or something is keeping openEvent around. The path is 'function n->Call->LexicalEnvironment->a->current->openEvent->FocusEvent'
My guess something somewhere is forgetting to clear openEvent, or that current.
Operating system
No response
Browser information
No response
URL for webapp
No response
Application version
No response
Homeserver
No response
Will you send logs?
No