Revert "deps: update libuv to 1.52.1"

[jasnell]

This reverts commit 5bebd7e.

The update appears to be causing persistent flakiness in test-http(s), test-net-, and test-tls- tests on Windows. The flakiness manifests as an abort apparently caused by a race condition introduced in connection cleanup on process teardown. Root cause still needs to be fully determined. Reverting the update will hopefully get us back to a stable state while the underlying issue is investigated and resolved.

[nodejs-github-bot]

Review requested:

• @nodejs/security-wg

[jasnell]

Note that i was able to repro the issue on windows locally; and after applying the revert locally it has not reoccurred yet.

[gurgunday]

lgtm

[anonrig]

cc @nodejs/libuv

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/72221/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.71%. Comparing base (e78ccd8) to head (8d72d50).
⚠️ Report is 18 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff            @@
##             main   #62497    +/-   ##
========================================
  Coverage   89.70%   89.71%            
========================================
  Files         691      692     +1     
  Lines      213935   214039   +104     
  Branches    41050    41065    +15     
========================================
+ Hits       191907   192015   +108     
+ Misses      14095    14092     -3     
+ Partials     7933     7932     -1     

see 53 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[lpinca]

@jasnell I see the issue on CI runs for #62165 and #62158 and that was one week before the libuv update landed, so I don't think the regression is in libuv. I'll try to run git-bisect if I find a way to reproduce, a Windows machine, and time.

[jasnell]

Hmmm.. weird. I haven't been able to reproduce locally on my machine here following the revert.. but then again, this is a really fast machine. If it's a race condition, I have to wonder if system performance has an impact. It failed about 1 out of 10 runs before the revert and 0 out of 10 after the revert. When i'm able to get back to it I'll run the tests longer to see if it shows up without the revert.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/72244/

[aduh95]

10 runs is probably too small a sample to be confident. Are you using the --repeat flag? python tools/test.py -t 9 --repeat 999 'test/parallel/test-http*' 'test/parallel/test-net-*' 'test/parallel/test-tls-*'

[lpinca]

I can reproduce locally

~/node ((2e5731eeebe...))
$ python ./tools/test.py --repeat=3000 parallel/test-net-server-keepalive
[06:19|% 100|+ 3000|-   0]: Done

All tests passed.

~/node ((5bebd7eaea2...))
$ python ./tools/test.py --repeat=3000 parallel/test-net-server-keepalive
=== release test-net-server-keepalive ===
Path: parallel/test-net-server-keepalive
Command: C:\Users\lpinca\node\out\Release\node.exe C:\Users\lpinca\node\test\parallel\test-net-server-keepalive.js


...


=== release test-net-server-keepalive ===
Path: parallel/test-net-server-keepalive
Command: C:\Users\lpinca\node\out\Release\node.exe C:\Users\lpinca\node\test\parallel\test-net-server-keepalive.js


[06:48|% 100|+ 2992|-   8]: Done

Failed tests:
C:\Users\lpinca\node\out\Release\node.exe C:\Users\lpinca\node\test\parallel\test-net-server-keepalive.js

...

C:\Users\lpinca\node\out\Release\node.exe C:\Users\lpinca\node\test\parallel\test-net-server-keepalive.js

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/72276/

[github-actions]

Fast-track has been requested by @panva. Please 👍 to approve.

[jasnell]

Don't merge yet. I ran tests last night and it still failed with the revert but at a much much lower rate. @mcollina is working on bisecting.

@aduh95, I didn't say I only ran it 10 times, I said it was failing 1 out of every 10 times. I haven't looked at the updated rate yet I only just saw that it still had some crashes. Need to test a bit more and I want to see what Matteo's bisect comes up with.

[aduh95]

Maybe that's what you meant, but not what you said x) it's interesting you found a 10% failure rate when Luigi found 0.3% in #62497 (comment); in any case, I think it's fine to land ASAP given the current state of CI

[jasnell]

What I've seen locally is the the rate appears to be rather variable. I tried it on a second windows machine with a much slower processor and the rate was significantly different... really just bolsters the case for it being some kind of race condition somewhere. In any case, let's not be in a rush to land this. I want to see what @mcollina's bisect turns up.

[lpinca]

With the case above, the issue starts with libuv/libuv@3a9a6e3

[juanarbol]

@jasnell can we close this in favor of #62511?

[santigimeno]

Tentative fix here: libuv/libuv#5107. Would love if someone that could locally reproduce this error could test it. Thanks

[lpinca]

@santigimeno I've

1. reverted 0474a27
2. applied https://patch-diff.githubusercontent.com/raw/libuv/libuv/pull/5107.diff

and it seems to work

~/node (main)
$ python ./tools/test.py --repeat=3000 parallel/test-net-server-keepalive
[03:04|% 100|+ 3000|-   0]: Done

All tests passed.

We can cherry-pick the fix when it lands. Thanks!

[santigimeno]

@lpinca thanks! The PR just landed in case you want to pick it up.