Fix: Windows development

[lukpsaxo]

Closes #7924

✅ Pull Request Checklist:

• Included link to corresponding React Spectrum GitHub Issue.
• Added/updated unit tests and storybook for this change (for new code or code which already has tests).
• Filled out test instructions.
• Updated documentation (if it already exists for this component).
• Looked at the Accessibility Practices for this feature - Aria Practices

📝 Test Instructions:

I know the core team is all on Mac - maybe check yarn test still works there.

Part of the fix taken from this suggestion: #2429 (comment)

🧢 Your Project:

Saxo Bank

[LFDanLu]

Thanks for the PR! I just tested it locally on my Windows machine and confirmed that the tests actually run now, but I seem to get a bunch of errors related missing translations due to the i18n files not being properly parsed or something. Pretty new to developing to Windows so I'll poke around and see if I patch a couple of things, but just wanted to know what configuration you are running on your end.

[lukpsaxo]

I only get tests working enough to test my other PR - but I'll see if the other failing tests are easy fixes.

[lukpsaxo]

Ok, I got it working better. It will now fail a little bit the first run and then be ok.

long winded explanation:

Unfortunately this code is unreliable in windows:

function writeCacheFile(cacheFile, res) {
  // If cache file already exists, return it.
  if (fs.existsSync(cacheFile)) {
    return cacheFile;
  }

  // Otherwise, write it atomically to avoid race conditions between threads/processes.
  let tmpFile = cacheFile + '.' + process.pid + (threadId != null ? '.' + threadId : '');
  fs.writeFileSync(tmpFile, res);
  fs.renameSync(tmpFile, cacheFile);
  return cacheFile;
}

There is a package - write-file-atomic - which is better but still fails in windows sometimes. In fact we have to have a patched version of jest to deal with this :/

jestjs/jest#11104 (comment)

and the only way to fix this 100% for windows is to make the resolver async or change the approach - for instance I would make this a transform and not done as part of the resolution, which would fix the problem.

But thats a bigger change. I can live with tests failing the first time.

[LFDanLu]

Thanks for the updates, confirmed that the tests are passing on the second run on windows and that things seems to run fine on MacOS. Will have to dig into patching writeCacheFile later but I think this is fine for now, will need to leave a note in the docs/FAQ

[lukpsaxo]

I will wait for:

#7933

to be merged

and then rebase this (unless you prefer to go ahead and merge and rebase #7933)

[lukpsaxo]

rebased - ready for review and merge

[lukpsaxo]

any chance of a second approval and merge?