deps: upgrade to libuv 1.42.0

[lpinca]

The first commit is from #39591.

Notable changes:

• win,tcp: make uv_close() work more like unix.
• cleanup,win: Remove _WIN32 guards on threadpool.
• more errno mappings and fixes.
• higher performance try-writes.
• fix string encoding issue of uv_os_gethostname() (note: MINGW-W64
  upstream is broken on i686 due to
  https://sourceforge.net/p/mingw-w64/bugs/899/).
• zOS support.
• Workarounds for a copy_file_range() kernel bug.
• Better support for TSan.
• darwin: use RLIMIT_STACK for fsevents pthread.
• fix a bug that would cause libuv to hang if the user called
  uv_shutdown().
• darwin: fix uv_barrier() race condition.

Fixes: #39502

[lpinca]

cc: @vtjnash

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39269/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39271/

[tniessen]

RSLGTM

[lpinca]

I need to help to update uv.gyp for z/OS. cc: @nodejs/build @richardlau?

[richardlau]

I need to help to update uv.gyp for z/OS. cc: @nodejs/build @richardlau?

We don't (can't?) currently build Node.js on z/OS in the CI so it shouldn't be a blocker to land this PR.
cc @nodejs/platform-zos

[lpinca]

We don't (can't?) currently build Node.js on z/OS in the CI so it shouldn't be a blocker to land this PR.

Should we remove or comment z/OS configuration in common.gypi and uv.gyp then? I find it confusing to have an outdated configuration that is not used.

[richardlau]

We don't (can't?) currently build Node.js on z/OS in the CI so it shouldn't be a blocker to land this PR.

Should we remove or comment z/OS configuration in uv.gyp then? I find it confusing to have an outdated configuration that is not used.

It wouldn't be the only not used configuration in uv.gyp (I think we still have Android and IOS in there).
The intent was to eventually build z/OS but I think we're still waiting for upstream patches to land here other than libuv (e.g. V8).

[lpinca]

Ok, I will revert the z/OS related changes that I have already applied to uv.gyp and keep it as is.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39274/

[vtjnash]

Note that you may need my #36111 PR also, as the commits in this libuv release were previously observed to exacerbate the CI issues with several TLS tests (due to an existing potential denial-of-service bug in the TLS stack, where nodejs may decide to abort if the https client crafts an RST packet)

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39286/

[cjihrig]

RSLGTM

[lpinca]

The test-asan workflow seems to consistently fail. parallel/test-cluster-primary-kill.js times out.

Edit: I could not reproduce locally. I've rebased against master and it passes now. Not sure if related.

[lpinca]

One test is flaky on macOS as per @vtjnash comment above:

=== release test-http2-buffersize ===
Path: parallel/test-http2-buffersize
Error: --- stderr ---
node:events:371
      throw er; // Unhandled 'error' event
      ^

Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:220:20)
Emitted 'error' event on ServerHttp2Stream instance at:
    at emitErrorNT (node:internal/streams/destroy:164:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)
    at processTicksAndRejections (node:internal/process/task_queues:83:21) {
  errno: -54,
  code: 'ECONNRESET',
  syscall: 'read'
}
Command: out/Release/node /Users/runner/work/node/node/test/parallel/test-http2-buffersize.js

[lpinca]

@Trott the second commit is basically a revert of c452632. See the commit message for details.

[lpinca]

I'll move it to a separate PR and rebase this on top of it.

[vtjnash]

Yes, that is one of the tests that had been known to run into the bug I mentioned as of last year (libuv/libuv#2943 (comment)). I am not sure why reverting a new commit back to the old state (which was previously known to fail without #36111) would make a significant difference to it?

[lpinca]

I am not sure why reverting a new commit back to the old state (which was previously known to fail without #36111) would make a significant difference to it?

With 9807960

$ ./tools/test.py -J --repeat=1000 test/parallel/test-http2-buffersize.js 
[00:10|% 100|+ 1000|-   0]: Done

Without

$ ./tools/test.py -J --repeat=1000 test/parallel/test-http2-buffersize.js 
=== release test-http2-buffersize ===                   
Path: parallel/test-http2-buffersize
node:events:371
      throw er; // Unhandled 'error' event
      ^

Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:220:20)
Emitted 'error' event on ServerHttp2Stream instance at:
    at emitErrorNT (node:internal/streams/destroy:164:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)
    at processTicksAndRejections (node:internal/process/task_queues:83:21) {
  errno: -54,
  code: 'ECONNRESET',
  syscall: 'read'
}
Command: out/Release/node /Users/luigi/code/node/test/parallel/test-http2-buffersize.js
=== release test-http2-buffersize ===                   
Path: parallel/test-http2-buffersize
node:events:371
      throw er; // Unhandled 'error' event
      ^

Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:220:20)
Emitted 'error' event on ServerHttp2Stream instance at:
    at emitErrorNT (node:internal/streams/destroy:164:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)
    at processTicksAndRejections (node:internal/process/task_queues:83:21) {
  errno: -54,
  code: 'ECONNRESET',
  syscall: 'read'
}
Command: out/Release/node /Users/luigi/code/node/test/parallel/test-http2-buffersize.js

...

=== release test-http2-buffersize ===                   
Path: parallel/test-http2-buffersize
node:events:371
      throw er; // Unhandled 'error' event
      ^

Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:220:20)
Emitted 'error' event on ServerHttp2Stream instance at:
    at emitErrorNT (node:internal/streams/destroy:164:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)
    at processTicksAndRejections (node:internal/process/task_queues:83:21) {
  errno: -54,
  code: 'ECONNRESET',
  syscall: 'read'
}
Command: out/Release/node /Users/luigi/code/node/test/parallel/test-http2-buffersize.js
[00:10|% 100|+ 621|- 379]: Done

The difference seems to be huge.

[vtjnash]

Ah, I didn't realize someone had recently deleted the part of that test that was previously catching that bug (a5d4a39)

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39366/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39369/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/39429/

[lpinca]

Landed in c61870c.