Additional lints

[TimWhiting]

Working on some additional lints:

• Finish riverpod_no_mutate_sync (don't mutate providers synchronously inside build of widget or providers).
• Avoid read on autoDispose / use container.listen on autoDispose
• use ref synchronously in widget callbacks (similar to BuildContext across async gaps)
• Avoid global provider container
• Finish riverpod_ref_escape_scope (don't pass WidgetRef between widgets, dont pass Ref except to a Notifier constructor or field initializer

[rrousselGit]

Epic!

[rrousselGit]

By the way, it'd be nice if the lints could be made compatible with the code-generator too.

Some of them should be fine, but I'm sure some will fail

[TimWhiting]

By the way, it'd be nice if the lints could be made compatible with the code-generator too.

Some of them should be fine, but I'm sure some will fail

Yes, I was thinking the same thing. I was playing with the code generator for the first time yesterday.

[TimWhiting]

I had to increase the timeout on the test. In CI it might be even slower. Is the slowness because of the custom_lint package, or because of how the lint rules are written do you think? There are probably some caching / other optimizations I should do with some of the lints, but I just wanted to know if it is partly the custom_lint package?

[rrousselGit]

analyzer & analyzer_plugin are slow in general. But it's worth investigating if you have the time

Although if you can, I think it'd be better to look into fixing the issue where custom_lint doesn't work in release mode. It's more important.
Then there's the symlink issue with custom_lint, which end-up analyzing the entire flutter repo if the user uses fmv

[rrousselGit]

About riverpod_no_mutate_sync, a case I tought of is ref.invalidate/ref.invalidateSelf

They probably should not be allowed either, using the same reasoning.

[TimWhiting]

About riverpod_no_mutate_sync, a case I tought of is ref.invalidate/ref.invalidateSelf

They probably should not be allowed either, using the same reasoning.

What about ref.refresh?

[TimWhiting]

I think this is getting too big for one PR. I've added items on the Linter issue for improving performance, and adding support for the codegen.

[TimWhiting]

I'll try to fix the build before merging though

[codecov]

Codecov Report

Merging #1584 (0816704) into master (dea35d2) will not change coverage.
The diff coverage is n/a.

❗ Current head 0816704 differs from pull request most recent head 3fdd124. Consider uploading reports for the commit 3fdd124 to get more accurate results

@@           Coverage Diff           @@
##           master    #1584   +/-   ##
=======================================
  Coverage   93.20%   93.20%           
=======================================
  Files          51       51           
  Lines        1929     1929           
=======================================
  Hits         1798     1798           
  Misses        131      131           

[TimWhiting]

@rrousselGit Ready for review. Compatibility with codegen, and figuring out how to speed up / optimize these lints will be next.

Tests pass locally, I'll look at the CI issue tomorrow

[rrousselGit]

Awesome thanks!