test: add tests for clearBuffer state machine

[captainsafia]

Checklist

• make -j8 test (UNIX), or vcbuild test nosign (Windows) passes
• tests and/or benchmarks are included
• commit message follows commit guidelines

Affected core subsystem(s)

stream, test

Description of change

This checks to see that clearBuffer appropriately decrements the
correct values in _writableState when clearBuffer is invoked in
end.

Closes#8687.

[bnoordhuis]

Someone from @nodejs/streams should do the final sign-off but LGTM insofar I'm able to judge that.

CI: https://ci.nodejs.org/job/node-test-pull-request/5077/

[bnoordhuis]

Tiny nit: no blank line here.

[captainsafia]

@bnoordhuis: Good catch on the blank line! Updated the code!

[mcollina]

can you please trigger the corked and bufferedRequestCount things using the normal API?

[captainsafia]

@mcollina: Thanks for the comment! I've tried to address your review.

I set corked use the cork function but I'm having a little bit of trouble setting the bufferedRequestCount using the API. From what I can say, it is incremented in writeOrBuffer which is called in write if the chunk provided is a valid chunk. For some strange reason, this bit of code doesn't produce the desired effect (incrementing bufferedRequestCount by 1 as done in writeOrBuffer).

writable.write('some-stuff', 'utf8', common.mustCall(() => {
assert.strictEqual(writable._writableState.bufferedRequestCount, 1);
}));

I'm trying to trace what's going on but before I invest too much time I figured I might ask you and see if I am missing an obvious way to do this.

[mcollina]

@captainsafia you need to have an async _write, and have write() being issued. Another alternative is to use cork() and uncork(). Let me know if you have any more questions, happy to chat on IRC if needed.

[captainsafia]

@mcollina: Thanks for the guidance! I'll try pursuing it on my own for a bit next week and get in touch with you on IRC if I'm not getting anywhere.

[Trott]

Bump! @captainsafia Are you still working on this?

[captainsafia]

Yep! I did some tinkering with it yesterday but I don't think I have anything commit worthy. I'll see if I can get something tomorrow or Monday. Sorry for leaving this hanging!

[captainsafia]

@mcollina: Couldn't catch you on IRC so here goes.

@captainsafia you need to have an async _write, and have write() being issued.

I've got a barebones _write = (chunk, encoding, cb) implemented in the tests here. Should write be issue inside that function...or?

Another alternative is to use cork() and uncork().

I'm using cork to update the corked state right now but from what I can tell it doesn't change bufferedRequestCount so it's seems like not a complete alternative.

[Fishrock123]

We should consider using a regular function here, as we make no guarentees that apply() or call() may not be called on this internally in the future.

(Assigning arrow functions as prototype functions is generally bad practice because those do not work.)

[mcollina]

bufferedRequestCount is incremented when a write is in progress, or if the stream is corked:

node/lib/_stream_writable.js

Lines 310 to 318 in f3cf8e9

	if(state.writing||state.corked){
	varlast=state.lastBufferedRequest;
	state.lastBufferedRequest=newWriteReq(chunk,encoding,cb);
	if(last){
	last.next=state.lastBufferedRequest;
	}else{
	state.bufferedRequest=state.lastBufferedRequest;
	}
	state.bufferedRequestCount+=1;

.

You should issue multiple writes synchronously after cork(), check bufferedRequestCount and then uncork().

Note that bufferedRequestCount makes most sense if _writev is implemented, see

node/lib/_stream_writable.js

Line 409 in f3cf8e9

varl=state.bufferedRequestCount;

. You should probably double the checks, one for a stream just with _write, and one with a stream with _writev.

[Trott]

Ping @captainsafia! Still working on this? (No rush; I'm just making sure it's not stalled.)

[captainsafia]

Sorry for staying delayed on this. I'd love to see this own through but I think it is a tad bit above my pay grade at the moment so I'll close it and give someone else the opportunity to tackle it with some of the insights provided above.

[mcollina]

@captainsafia may I build upon your commit?

[captainsafia]

@mcollina: Sure! I've added you as a collaborator on my fork.

[mcollina]

@captainsafia I force pushed this. Let me know if you are ok with the content.

[mcollina]

@italoacasas should be addressed.

[italoacasas]

CI: https://ci.nodejs.org/job/node-test-pull-request/5664/console

[Trott]

CI failures on AIX and FreeBSD seem unrelated. I just terminated some stalled tests on those machines--probably a previous test run had a bug that caused cluster tests to hang and not free ports, causing subsequent runs like this one to also fail.

Re-running CI: https://ci.nodejs.org/job/node-test-pull-request/5665/

[italoacasas]

Landed 275362a

[captainsafia]

Thanks for getting this through the finish line, everyone! ❤️