Fallback straight to JS if realpath.native/realpathSync.native throw an exception

[rvaidya]

if fs.realpath.native and fs.realpathSync.native crash, it could be because we're on a weird configuration like a RAMdisk on Windows. if so, fall straight back to default JS implementation.

Otherwise, the process.binding attempt may cause the node runtime to abort

I also refactored the code to reduce repetition. Let me know if that is ok.

I was unable to run test.js - getting errors doing npm install coming from "babel-helper-regex" and then getting "Cannot find module 'eslint-config-simenb-node'" when trying to run it anyways.

I tested by copying my updated index.js into my project folder under jest-config and running it that way.

This also fixes a potential typo: both realpath and realpathSync were calling realpath in the process.binding attempt, which was probably unintentional.

[rvaidya]

#35

[rvaidya]

Getting issues from commitlint, but those can probably be addressed when you squash the commit.

[SimenB]

Thanks for the PR!

This also fixes a potential typo: both realpath and realpathSync were calling realpath in the process.binding attempt, which was probably unintentional.

No, that's intentional - there's no process.binding('fs').realpathSync - the function is always sync. Could you fix that?

$ node -p "Object.keys(process.binding('fs')).sort()"
[ 'FSReqWrap',
  'FileHandle',
  'StatWatcher',
  'access',
  'bigintStatValues',
  'chmod',
  'chown',
  'close',
  'copyFile',
  'fchmod',
  'fchown',
  'fdatasync',
  'fstat',
  'fsync',
  'ftruncate',
  'futimes',
  'internalModuleReadJSON',
  'internalModuleStat',
  'kFsStatsFieldsLength',
  'kUsePromises',
  'lchown',
  'link',
  'lstat',
  'mkdir',
  'mkdtemp',
  'open',
  'openFileHandle',
  'read',
  'readdir',
  'readlink',
  'realpath',
  'rename',
  'rmdir',
  'stat',
  'statValues',
  'symlink',
  'unlink',
  'utimes',
  'writeBuffer',
  'writeBuffers',
  'writeString' ]

Getting issues from commitlint, but those can probably be addressed when you squash the commit.

Yup, don't worry about those 🙂

[rvaidya]

updated

[rvaidya]

Sorry for not getting to this for so long.

I'm still not finding a real clean solution for this - since the async versions are using promises, the only way to handle errors in the first call is to chain promises. If if (typeof fs.realpath.native === 'function') is false, however, the fsBinding approach will still cause the runtime to abort. No great way to preemptively detect the specific Windows + RAMdisk situation and handle it across the board.

I have a callback chaining implementation at https://github.com/rvaidya/realpath-native/tree/feature/windowsRamdiskTry2

This commit doesn't try to refactor anything

[SimenB]

@rvaidya your solution in the linked branch seems OK to me, even though it turned out to be a bit of code. Wanna update this PR to that approach?

[rvaidya]

I opened #39 - you can close this PR

[SimenB]

I've published a v3 and deprecated it - all current releases of node supports fs.realpath.native now, so this module is no longer needed.

Jest 26 no longer depends on this module.