[Bug] False positive "potential listener LEAK detected"

Reproducible in vscode.dev or in VS Code Desktop?

• Not reproducible in vscode.dev or VS Code Desktop

Reproducible in the monaco editor playground?

• Not reproducible in the monaco editor playground

Monaco Editor Playground Link

Monaco playground link

Monaco Editor Playground Code

The key part is just adding this:


for (let i = 0; i < 200; ++i)
{
	monaco.editor.createModel("");
}

Reproduction Steps

Suppose the user has a project with 200 files in it. For this use case, a model must be created for every script file for features like 'Rename' and 'Find all references' to work correctly. It is not possible to lazy-create or dispose inactive models in this case: they must all be created simultaneously for such cross-file features to work as expected.

VS Code can open a project with 200 files just fine. However if you create 200 models with Monaco, it fails.

Actual (Problematic) Behavior

Throws an error Uncaught Error: [001] potential listener LEAK detected, having 200 listeners already.

Expected Behavior

Not to throw an error, or for the limit of 200 to at least be configurable.

Additional Context

This was previously filed at #3895 but it was closed without being resolved we just ran in to it again when integrating Monaco in to Construct. The previous comment was "that shouldn't happen in real applications", but as I've described, it really does happen and is a legitimate use case that VS Code can handle fine and only Monaco has an issue with.

The limit appears to come from leakWarningThreshold in the source code, but this limit is not configurable. For the time being we have to edit the source code to make the number 200 bigger. If there was at least an API to specify that, we wouldn't have to edit the source in our distribution.

Reproduce the issue from the linked Monaco playground with 200 calls to monaco.editor.createModel(""). Start by locating leakWarningThreshold and the listener-registration path that emits the warning. Done means the valid multi-model use case no longer throws, or the threshold can be configured through a supported API.