Node.js 16 failures

[targos]

I started a run (with citgm's main branch) against Node.js master to check the current status: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2643/

[targos]

New run: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/

@devinivy It looks like hapi is going to have issues: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/_hapi_hapi_v20_1_1/

@richardlau node-report seems a bit too strict in its tests for version numbers: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/node_report_v2_2_11/

@nodejs/npm: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/npm_v7_6_3/

@alexlamsl uglify-js is failing because of a spec change: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/uglify_js_v3_13_1/
See tc39/ecma262#2256

[devinivy]

We published hapi 20.1.2, and it should contain fixes for those issues 👍

[richardlau]

New run: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/

@richardlau node-report seems a bit too strict in its tests for version numbers: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2650/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/node_report_v2_2_11/

node-report is comparing the versions of Node.js dependencies in the report to process.versions -- the failure is occurring since we use each value in process.versions to construct a regular expression and the version string for OpenSSL for the fork we're now using contains a + character which needs to be escaped for use in a regular expression. Raised nodejs/node-report#147 to fix.

[MylesBorins]

@targos the npm failure seems to be related to the pre-release version of Node.js causing weird issues with the npm install phase.

[BethGriggs]

CITGM isn't particularly healthy on the v16.0.0 proposal (132 failures). v16.0.0 has not diverged from master yet (just slightly behind), so I suspect runs on master yield similar results.

Run on rc.2 (3bc20dc) - https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2671/testReport/

Probably best to start by capturing the failures that are occurring on many platforms:

Old failures

• @hapi/hapi-v20.1.2
  • Error: The following leaks were detected:atob, btoa, performance
• @hapi/shot-v5.0.5
  • Error: The following leaks were detected:atob, btoa, performance
• nan-v2.14.2
  • Link
    • tests: adapt for v8 8.9 nan#910
• node-report-v2.2.11
  • Now skipped, but waiting on CITGM publish.
• npm-v7.8.0
  • Error: Error: Cannot find module './reports/base'
• socket.io-v4.0.1
  • socket.io socket should not fire events more than once after manually reconnecting:
• uglify-js-v3.13.3
  • Link
    • Likely spec change mentioned above? Normative: forbid for-of loops with variable named async tc39/ecma262#2256
    • Issue raised Latest v8 breakages from Node.js v16 mishoo/UglifyJS#4846
• undici-v3.3.3
  • Link
• yargs-v16.2.0
  • Link
• yeoman-generator-v5.2.0'
  • Link

[richardlau]

* [ ]  undici-v3.3.3
  
  * [Link](https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/2671/nodes=debian9-64/testReport/(root)/citgm/undici_v3_3_3/)

At least one of the failures,

     # Subtest: 20 times GET with pipelining 10
         1..61
         not ok 1 - (unnamed test)
           ---
           code: DEP0111
           ...

Is the process.binding deprecation which was removed from undici in nodejs/undici#564 but hasn't been released.
Refs: nodejs/node#37813 (comment).

[BethGriggs]

I can reproduce the npm-v7.8.0 failures locally on the v16.0.0-proposal build 🤔

➜  cli git:(latest) ✗ ~/node/out/Release/node --version
v16.0.0
➜  cli git:(latest) ✗ ~/node/out/Release/node ~/cli/node_modules/tap/bin/run.js
Error: Cannot find module './reports/base'
Require stack:
- /Users/bgriggs/cli/node_modules/jackspeak/noop.js
    at Function.Module._resolveFilename (node:internal/modules/cjs/loader:940:15)
    at resolveFileName (/Users/bgriggs/cli/node_modules/tap/node_modules/resolve-from/index.js:17:39)
    at resolveFrom (/Users/bgriggs/cli/node_modules/tap/node_modules/resolve-from/index.js:31:9)
    at module.exports (/Users/bgriggs/cli/node_modules/tap/node_modules/resolve-from/index.js:34:41)
    at importJSX (/Users/bgriggs/cli/node_modules/tap/node_modules/import-jsx/index.js:24:21)
    at module.exports (/Users/bgriggs/cli/node_modules/tap/node_modules/treport/lib/index.js:13:18)
    at exports.makeReporter (/Users/bgriggs/cli/node_modules/tap/bin/run.js:422:23)
    at runTests (/Users/bgriggs/cli/node_modules/tap/bin/run.js:744:3)
    at mainAsync (/Users/bgriggs/cli/node_modules/tap/bin/run.js:244:5)
    at main (/Users/bgriggs/cli/node_modules/tap/bin/run.js:127:3) {
  code: 'MODULE_NOT_FOUND',
  requireStack: [ '/Users/bgriggs/cli/node_modules/jackspeak/noop.js' ]
}

My git bisect is pointing to the V8 9.0 upgrade nodejs/node#37587 - as the last known good commit to build/pass npm test appears to be the previous commit, 802b3e7cf9. But I haven't yet convinced myself that #37587 is responsible (or why).

/cc @nodejs/npm

[ruyadorno]

hi @BethGriggs! That is looking more like a https://github.com/tapjs/node-tap (which is the test framework we're using) problem than a npm one, we'll dig into it!