Roll back micromatch

[Tvrqvoise]

Summary

As part of #6400, micromatch was updated. While micromatch's CHANGELOG claims that this is safe, several regressions have been noted in cases where users relied upon invalid glob patterns. For example, these patterns would all match src/foo/bar/baz.js, but no longer do:

• src/**/*.{js}
• src/**.js
• src/**/*.{js|ts}

Fixes #6563
Fixes #6546

Test plan

Unit tests were added which demonstrate the known cases.

[facebook-github-bot]

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have you on file. In order for us to review and merge your code, please sign up at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need the corporate CLA signed.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Facebook open source project. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Facebook open source project. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Facebook open source project. Thanks!

[Tvrqvoise]

I don't understand why "test-node-8" is failing. It looks like a test within jest-diff has a leak, but I did not touch jest-diff. Can someone with more permissions either re-run the job or take a look?

[codecov-io]

Codecov Report

Merging #6661 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #6661   +/-   ##
=======================================
  Coverage   63.73%   63.73%           
=======================================
  Files         235      235           
  Lines        8940     8940           
  Branches        4        4           
=======================================
  Hits         5698     5698           
  Misses       3241     3241           
  Partials        1        1

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fa92643...7bda341. Read the comment docs.

[SimenB]

@mjesun @cpojer I think this is worth a patch release

[mjesun]

I’ll take care tomorrow of it 🙂

[davidleacy]

Definitely appreciate a patch release for this, currently blocking package updates for one of our projects.

[mjesun]

We'll do it. I'm waiting for some PRs to get merged and we'll perform a minor.

[jonschlinkert]

Hey folks, I'm the author of micromatch, just wanted to let you know that we'll be looking into the issue to try to understand what's happening. Based on the information provided by @Tvrqvoise, there shouldn't be any issues with those glob patterns.

While we look into it, my hunch is that it's one of the following:

1. an option needs to be set (or unset), such as unixify which ensures file paths use forward slashes on windows.
2. Node's path module is being used to join glob patterns or make them absolute, etc, and something isn't being joined correctly.

In the meantime, if anyone has any more info they can share, please feel free to create an issue on the micromatch repository so we can get to the bottom of this and figure it out. thanks!

[rickhanlonii]

Thanks @jonschlinkert!

[Tvrqvoise]

@jonschlinkert, I can see differing behavior for the test cases I added. I'm on OSX High Sierra, for what it matters. I'll throw together a minimal test case on your repository's issues page tomorrow.

[jonschlinkert]

I can see differing behavior for the test cases I added in the unit test. I'm on OSX High Sierra, for what it matters. I'll throw together a minimal test case on your repository tomorrow.

Thanks!

[jonschlinkert]

I didn't catch this before, but all three of the patterns mentioned by @Tvrqvoise in the OP are invalid globs. (since I didn't catch it, and I spend a lot of time looking at globs, it's obviously... um, not obvious).

I go into more detail here.

[SimenB]

Thanks for taking a look! I think the changes in v3 are all fine, but I also think holding off on the upgrade until the next major of Jest makes sense - it broke a lot of people's setup. Even though it broke because of invalid globs I think this should come in a major, along with these exact examples instructing people on how to migrate :)

[jonschlinkert]

Even though it broke because of invalid globs I think this should come in a major, along with these exact examples instructing people on how to migrate :)

Agreed! In the meantime I'm happy to help if anyone has issues.

Regarding the Windows path issue, in case it helps anyone for now, on our own projects we use to-absolute-glob to, um, ...create absolute globs (I seem to be doing this a lot. I must be stuck in some kind of language recursion loop lol).

[SimenB]

@jonschlinkert do you by any chance know of a module that can validate the globs? We could print a deprecation warning for Jest 23 to help people fix their globs in time for jest 24

[jonschlinkert]

@SimenB that's an interesting question. Let me think about that, and how it would work. I might have a couple of ideas. Off-hand, it seems like a catch-22, meaning that a validator couldn't be trusted for the same reason that we can't say for sure if the user intended to use backslashes as escape characters versus path separators. The only way to know for sure is to have the path before it's joined to the glob.

However, it's quite possible that it's much simpler and I'm hung up on something that isn't important. Have there been any issues reported by users related to this yet? If I could see specific scenarios it might help me understand your use case more thoroughly and come up with ideas. Either way I'd be happy to help.

[SimenB]

Escaping vs path separator is one thing, I'm thinking more of the simple cases such as the ones mentioned in the OP in this PR.

It doesn't have to be perfect, but if we can flag some obviously wrong ones that'd be awesome, and waaay better than nothing 🙂

[jonschlinkert]

Ah, got it. I might need some clarification then, since my understanding was that this PR was user-error since the patterns listed are all invalid glob patterns (which we didn't notice at first).

Fwiw, we haven't had any other issues related to matching or regressions on micromatch itself. Not that issues can't surface here or elsewhere, but IMHO this issue here wasn't really an issue, as it was a false-negative.

Perhaps this comment will shed some light on that point.

edit: btw my intention isn't to be dismissive of the issue, I'm still very happy to help with whatever resolution is needed. Also fwiw, I mention in that comment about "being more accurate", but we haven't had any other issues about it besides that one.

[SimenB]

my understanding was that this PR was user-error since the patterns listed are all invalid glob patterns (which we didn't notice at first).

Yes exactly. What I meant is that it would be great if we could say "hey, these globs are invalid, please fix them as they will stop working" for user-supplied patterns

[juliancoleman]

Is there any word on this? Tests still fail on Windows with jest@^23.5.0

[SimenB]

@juliancoleman micromatch has been rolled back in that version, so id you still have issues I suggest opening up a new issue 🙂

[github-actions]

This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.
Please note this issue tracker is not a help forum. We recommend using StackOverflow or our discord channel for questions.