-
Notifications
You must be signed in to change notification settings - Fork 125
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
Run CI against minimum supported Sorbet version #1252
base: main
Are you sure you want to change the base?
Conversation
gemfiles/Gemfile-sorbet-minimum
Outdated
@@ -0,0 +1,5 @@ | |||
# frozen_string_literal: true | |||
|
|||
gem("sorbet-static-and-runtime", "0.5.9204") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As noted in my description, this matches the minimum version we specify in tapioca.gemspec
(">= 0.5.9204"
).
- This version does not exist.
We should either switch to a version number that does exist, or fallback to specifying Sorbet dependencies individually. - Ideally we'd do this dynamically so this file doesn't need to be kept in sync.
Perhaps simplest to just do string matching on theGemfile
and extract the version?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For 1. I think we should first create a PR that bumps the version to an existing one. The first version for sorbet-static-and-runtime
is v0.5.9519 which is almost a year old. We can bump to something more recent.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've opened #1260 which starts by bumping to the first version.
What version do you think we should use as the requirement? We could go with the latest, but I'm not sure if we want to force the update for consumers.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's keep the one you used in #1260 for the time being 👍
Thanks!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To tie up loose ends, #1260 ended up bumping us to the first version that allows us to remove our feature flag checks (because all features are supported).
spec.add_dependency("sorbet-static-and-runtime", ">= 0.5.9892")
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Okay, I had to nudge it forward again to 0.5.9896
, because 0.5.9892
doesn't actually exist either 😅
@@ -21,7 +21,7 @@ group(:development, :test) do | |||
gem("smart_properties", require: false) | |||
gem("frozen_record", require: false) | |||
gem("sprockets", require: false) | |||
gem("rails", require: false) | |||
gem("rails", require: false) unless defined?(@specified_rails) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not convinced on the need to pollute the default Gemfile just for the sake of avoiding duplication in the gemfiles used for tests.
What do you think about creating an ERB template for the test gemfiles that could be parameterized based on the Rails version we want to test and the sorbet version parsed from the gemspec?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The thing is that if we do that, we still end up duplicating the rest of the Gemfile, which leads to it becoming out of sync (as it currently is on main
).
I based this approach off the approach taken by Shopify/maintenance_tasks, although I used instance variables differently due to differences in the arguments to gem
(specifically, the presence of require: false
).
Specifically with regards to using a template, that would mean we'd have to have a step in CI that renders the template before bundle install
reads it, but that would necessarily mean doing it before installing any gems, which might be problematic.
Honestly, the cleanest solution would be if we were able to do something like
eval_gemfile "../Gemfile" # Containing a regular gemfile
override do
# No complaint about duplicate gem specification
gem("rails", "~> 6.1.0", require: false)
end
But AFAIK Bundler doesn't support anything like that.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IMO removing that duplication (and the chance for them to get out of sync) justifies the change here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not saying this is better, but in this project I use env var to control which version to use (from Rails 5 to main).
Pros:
- We only need 1 Gemfile and logics are concentrated
- No warnings AFAICT
Cons:
- If conditions
The main Gemfile specifies 0.3.2, but the other two specify 0.3.1.
Rather than duplicate the entire file, we can use eval_gemfile.
ec00d0a
to
7e081ce
Compare
0.5.9896 is the first published version that is actually >= 0.5.9892
Extract minimum Sorbet version from gemspec and run against that to make sure we're actually compatible.
7e081ce
to
1937577
Compare
@@ -21,7 +21,7 @@ group(:development, :test) do | |||
gem("smart_properties", require: false) | |||
gem("frozen_record", require: false) | |||
gem("sprockets", require: false) | |||
gem("rails", require: false) | |||
gem("rails", require: false) unless defined?(@specified_rails) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IMO removing that duplication (and the chance for them to get out of sync) justifies the change here.
Motivation
In #1046 (comment), there was a proposal to use a syntax that would not have been supported on the version of Sorbet we list as our minimum. It occurred to me that something in CI should exist to catch that sort of thing.
Therefore, this PR's goal is to extend the CI matrix to run against the minimum version of Sorbet which we claim to support. This way, we're able to detect incompatibilities as they are introduced, and either alter the implementation or version requirement, keeping the
gemspec
accurate.Implementation
sorbet-static-and-runtime
matching the minimum version we list ingemspec
.Tests
No new tests should need to be introduced, but this branch actually already discovers that the version we specify as our minimum doesn't actually exist.
I suspect we were previously specifying matching versions of
sorbet
,sorbet-static
, andsorbet-runtime
, which we then migrated tosorbet-static-and-runtime
(keeping the same version number), without realizing the version number started later than our "current" version.