esm: use sync loading/resolving on non-loader-hook thread

[joyeecheung]

esm: use sync loading/resolving on non-loader-hook thread

ESM resolution and loading is now always synchronous from a
non-loader-hook thread. If no asynchrnous loader hooks are
registered, the resolution/loading is entirely synchronous.
If asynchronous loader hooks are registered, these would be
synchronous on the non-loader-hook thread, and asynchronous
on the loader hook thread.

This avoids several races caused by async/sync loading sharing
the same cache. In particular, asynchronous loader hooks
now works with require(esm) - previously it tends to break
due to races.

In addition, when an asynchronous loader hook
returns a promise that never settles, the main thread no longer
silently exits with exit code 13, leaving the code below
any module loading calls silently ignored without being executed.
Instead, it now throws ERR_ASYNC_LOADER_REQUEST_NEVER_SETTLED
which can be caught and handled by the main thread. If the module
request comes from import(), the never-settling promise is
now relayed to the result returned by import().

Drive-by: when annotating the error about importing undetectable
named exports from CommonJS, it now no longer reload the source
code of the CommonJS module, and instead reuses format information
cached when the module was loaded for linking.

Fixes: #59666

[nodejs-github-bot]

Review requested:

• @nodejs/loaders
• @nodejs/test_runner

[codecov]

Codecov Report

❌ Patch coverage is 94.57831% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.57%. Comparing base (b19525a) to head (f9deee5).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/internal/modules/esm/loader.js	91.90%	14 Missing ⚠️
lib/internal/modules/esm/module_job.js	97.43%	3 Missing ⚠️
lib/internal/modules/esm/hooks.js	94.11%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #60380      +/-   ##
==========================================
- Coverage   88.57%   88.57%   -0.01%     
==========================================
  Files         704      704              
  Lines      208444   207753     -691     
  Branches    40057    40013      -44     
==========================================
- Hits       184627   184014     -613     
+ Misses      15848    15786      -62     
+ Partials     7969     7953      -16     

Files with missing lines	Coverage Δ
lib/internal/errors.js	97.51% <100.00%> (+<0.01%)	⬆️
lib/internal/modules/esm/initialize_import_meta.js	100.00% <100.00%> (ø)
lib/internal/modules/esm/translators.js	92.93% <100.00%> (+0.03%)	⬆️
lib/internal/modules/esm/utils.js	100.00% <100.00%> (ø)
lib/internal/test_runner/mock/mock.js	98.72% <100.00%> (-0.01%)	⬇️
lib/internal/modules/esm/hooks.js	86.24% <94.11%> (-5.66%)	⬇️
lib/internal/modules/esm/module_job.js	96.59% <97.43%> (-0.18%)	⬇️
lib/internal/modules/esm/loader.js	96.84% <91.90%> (+1.05%)	⬆️

... and 112 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.

[GeoffreyBooth]

I think this gets us a long way toward completing #55782? Like for example I think this means we can default to the ESM loader now if we want to (#50356). If so, this is a huge improvement!

Not to say that we necessarily want to switch the default loader anytime soon if ever, just that I assume we're no longer blocked from doing so by the async hooks tests.

[joyeecheung]

Like for example I think this means we can default to the ESM loader now if we want to

I think that's a separate thing - the triggered hooks you see likely comes from the async activity from the outer layer of ModuleLoader.import, this PR only refactors some inner paths of it, it won't remove the async activity from the outer layer. There might also be other unwanted consequences of using the ESM loader to load a CJS entry point (e.g. overhead from lexing and creating a module facade, which is usually unnecessarily for an entrypoint, but needs to be done in case some other module comes back loading the entry point like a module, which is sometimes what people use require.main for)

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/69869/

[saico-0ps]

Thanks

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/69870/