stream: eos premature close

[ronag]

For writable premature close is close before finish, i.e. it should look for finished instead of ended. Would be better with a finishEmitted but we don't have that.

For readable we should be checking endEmitted (which we do first but not later).

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]

I'm unsure how to make tests that actually trigger this... but the semantics I think are more correct like this

[ronag]

This actually fixes a bug (close before end should be an error). Though it should probably be sem-major due to the updated test.

[ronag]

@benjamingr who's a good ping on this one?

[benjamingr]

I don't mind reviewing it - can you explain the tradeoffs in the new behavior vs. the existing one?

[ronag]

@benjamincburnsclose before finish (writable) and close before end (readable) should be premature close error. Which it isn't in every case before this PR (see test update).

[benjamingr]

@nodejs/streams

[benjamingr]

LGTM with a green citgm run

[mcollina]

Marking it as semver-major as it changes behavior.

[mcollina]

What is the bug this is fixing? What’s the goal of this change?

[ronag]

@mcollina

What is the bug this is fixing? What’s the goal of this change?

close before finish (writable) and close before end (readable) should be premature close error. Which it isn't in every case before this PR (see test update).

[benjamingr]

@ronag just to elaborate on Matteo's comment regarding semverness - changes to the order of the events streams (or in general node APIs) fire are pretty substantial changes with possible big implications on the ecosystem so generally they're all semver-major (means they'll be released in the next major release - in this case 13)

[ronag]

@benjamingr no problem, makes perfect sense

[ronag]

updated test

[ronag]

merged into #28748