src,http2: ensure cleanup if a frame is not sent

[ywave620]

Fix 41582
Call to JS and close the session if a frame is not sent even there is no frameError listener registered by user.

[nodejs-github-bot]

Review requested:

• @nodejs/http2
• @nodejs/net

[ywave620]

The evidence of the absence of 'frameError' listener is the blame is that if you add such a listener to the http2Session in the server side in the test, client can get a response.
This PR ensures 'onFrameError' in JS, which takes care of cleanup get called even if no user registered 'frameError' listener.

node/lib/internal/http2/core.js

Lines 602 to 615 in 8e203f1

	// Called by the native layer when an error has occurred sending a
	// frame. This should be exceedingly rare.
	function onFrameError(id, type, code) {
	const session = this[kOwner];
	if (session.destroyed)
	return;
	debugSessionObj(session, 'error sending frame type %d on stream %d, code: %d',
	type, id, code);
	const emitter = session[kState].streams.get(id) || session;
	emitter[kUpdateTimer]();
	emitter.emit('frameError', type, code, id);
	session[kState].streams.get(id).close(code);
	session.close();
	}

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/50590/

[mcollina]

lgtm

[ywave620]

Don't know what was going on, many tests ended up with no report generated, likes

Could anyone re-run the CI?

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/50606/

[ywave620]

@mcollina are we ready to merge this? :)

[RafaelGSS]

I was afraid it would revert the memory leak fix I did a while ago. But, the since the CI is green I think we are good. LGTM.

[mcollina]

@RafaelGSS could you reopen that issue?

[nodejs-github-bot]

Landed in 18e1f3c

[RafaelGSS]

@RafaelGSS could you reopen that issue?

Why? I meant I didn't approve this PR before because I was waiting for the CI to run to guarantee this won't bring back the memory leak issue.