stream: ensure errorEmitted is always set

[ronag]

_writableState.errorEmitted should always be set on error.

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

[ronag]

@mcollina ping (sorry for spam?)

[mcollina]

Would you mind adding another unit test on the actual behavior change? What consequences has setting errorEmitted in there?

[ronag]

@mcollina: Not sure what you expect there :/

[mcollina]

I don't know why you are proposing this change. What's the end user result for this?

[ronag]

@mcollina: Because sometimes errorEmitted is set and sometimes it is not with no clear reason. Why would you not do this change to ensure errorEmitted is (almost) always what you assume it to be?

e.g. _writable_stream:

SET

https://github.com/nodejs/node/blob/master/lib/_stream_writable.js#L431
https://github.com/nodejs/node/blob/master/lib/_stream_writable.js#L438

NOT SET

https://github.com/nodejs/node/blob/master/lib/_stream_writable.js#L252
https://github.com/nodejs/node/blob/master/lib/_stream_writable.js#L268
https://github.com/nodejs/node/blob/master/lib/_stream_writable.js#L618

[mcollina]

+1.

I prefer tests that try to double-emit errors and we block them rather than tests about internals. That's what I'm asking for.

Why would you not do this change to ensure errorEmitted is (almost) always what you assume it to be?

I'm conservative of changes, as every change we do to streams can break a lot of people.

[ronag]

I prefer tests that try to double-emit errors and we block them rather than tests about internals. That's what I'm asking for.

Thank you. Makes sense.

I'm conservative of changes, as every change we do to streams can break a lot of people.

Haha, I guess that is your job.

I'm very keen on consistency and predicability. Unfortunately at the moment it is almost a requirement to read the Node code to figure out how things actually work before using them. I will try to help out within your zone.

[mscdex]

I agree, we should be running CITGM for the event and stream-related changes in this and the other PRs.

[ronag]

@mcollina tests updated

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/24563/

[mcollina]

LGTM

[mcollina]

CI: https://ci.nodejs.org/job/node-test-pull-request/24880/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/24880/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/24890/

[mcollina]

CITGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/1937/

[mcollina]

Can you rebase on top of master? CITGM is failing because of that esm bug.

[ronag]

Rebased.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/24903/

[mcollina]

CITGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/1941

[ronag]

This was fixed in 4a2bd69