Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

update postcss to version 7 #231

Closed
wants to merge 1 commit into from
Closed

Conversation

EECOLOR
Copy link

@EECOLOR EECOLOR commented Oct 23, 2018

please note that you need to set the correct version and names for:

  • postcss-modules-extract-imports
  • postcss-modules-local-by-default
  • postcss-modules-scope
  • postcss-modules-values

The tests are failing. In some case it might be because they reference a local version of the dependency. In others the behavior of the dependency is changed. I lack domain knowledge to determine if the changes are ok.

css-modules/postcss-modules-extract-imports#150
css-modules/postcss-icss-selectors#125
css-modules/postcss-icss-composes#169
css-modules/postcss-icss-values#113

@EECOLOR
Copy link
Author

EECOLOR commented Oct 23, 2018

Also adjust travis build. Node 4 is no longer supported: https://github.com/postcss/postcss/releases/tag/7.0.0

@lukeed
Copy link

lukeed commented Jun 20, 2020

This is so necessary.

This module – on its own – pulls in two distinct versions of postcss: graph

It's directly & solely responsible for postcss@6.0.1 being loaded.
It shares responsibility for postcss@6.0.23 with:

  • postcss-modules-scope@1.1.0 – (2.2.0 available)
  • postcss-modules-local-by-default@1.2.0 – (3.0.2 available)
  • postcss-modules-values@1.3.0 – (3.0.0 available)
  • postcss-modules-extract-imports@1.1.0 – (2.0.0 available)

Then postcss-modules, a popular dependent of this package, requires postcss@7.0.32 – as it should, since that's the latest.

In any event postcss-modules loads 49 packages total in its network.
This css-modules-loader-core package is responsible for 38 of them.

No wonder postcss-modules takes so long to install.

If you don't care about this package anymore, please transfer it to someone who does.

"postcss-modules-local-by-default": "1.2.0",
"postcss-modules-scope": "1.1.0",
"postcss-modules-values": "1.3.0"
"postcss": "7.0.5",
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I know this is not EECOLOR's fault – the PR is 1.5 years old – but pinning to 7.0.5 would cause the same issue I've pointed out.

This should have always been a semver range.

Suggested change
"postcss": "7.0.5",
"postcss": "^7.0.0",

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As a side note, some versions before 7.0.5 contained bugs that prevented stuff from working. It's too long ago to remember which ones. If you still want to go this route I would suggest ^7.0.5

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It does not matter. By specifying a semver range, any dependents will resolve the latest 7.x match.

In this case, everyone will be downloading 7.0.32 as that's currently the latest.
Pinning a specific version is a large part of what got poatcss-modules into this mess.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

While you are right in most cases, if a user has 7.0.3 for example and he adds this module, things will break.

That said, practically I think it's fine to use ^7.0.0 because 7.0.5 was released long enough ago. The chance a user has a lower than 7.0.5 is very slim.

For these types of libraries I think I would even move the dependency to the peerDependencies.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Agreed, although I'm not sure if yarn@2 and pnpm provision peerDep access correctly. I assume they do, but not 100% sure.

In your example, if the user already had 7.0.3 locally, they'd at least keep that. They certainly wouldn't do back down to 7.0.0. But in 99% of cases, the fresh install of this module (assuming it ever gets out) triggers a refresh of its dependencies, resolving to the latest version that matches the specified range.

@EECOLOR
Copy link
Author

EECOLOR commented Jun 20, 2020

@lukeed
Copy link

lukeed commented Jun 20, 2020

I'm just looking to get things fixed as quickly and simply as possible. I don't think it's wise to hold out/dream for plugin rewrites when something like dependency resolution has been a blocker/unmaintained for ~2 years

@EECOLOR
Copy link
Author

EECOLOR commented Jun 20, 2020

That is exactly the reason why we stepped away from using this module and just used the separate underlying modules that do the actual work.

@jquense
Copy link

jquense commented Dec 1, 2020

who holds the keys to this package? @alexander-akait do you have access to this? most bundle css-modules support outside of webpack is using this incredibly outdated tool. It deserves love. I would be happy to help update and maintain it if someone who has access could please hand over the keys 🙏 css-modules is too pervasive to be abandonware and i'm happy to help! cc @markdalgleish @geelen

@EECOLOR
Copy link
Author

EECOLOR commented Dec 1, 2020

@jquense In hindsight using this specific module proved to cause problems. That is why we now embrace the move to the underlying libraries. If you do want to try and create a package like this, you might find some inspiration here:

https://github.com/kaliberjs/build/blob/master/library/webpack-loaders/css-loader.js
https://github.com/kaliberjs/build/blob/master/library/postcss-plugins/postcss-import-export-parser.js

@alexander-akait
Copy link
Member

How I can help?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants