Remove ruby platform check for sorbet dependency

[TonyCTHsu]

What does this PR do?

With the ruby platform check, the appraisal lockfile would always end up with difference. However, currently we are using a compatible version that could be installed on x86 or arm64-darwin21 from M1 Mac. So we could remove this conditional

How to test this

With M1 Mac, bundle install && bundle exec rake typecheck

[codecov-commenter]

Codecov Report

Merging #2445 (66fd91a) into master (970c87a) will decrease coverage by 0.00%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #2445      +/-   ##
==========================================
- Coverage   98.34%   98.33%   -0.01%     
==========================================
  Files        1101     1101              
  Lines       58977    58974       -3     
==========================================
- Hits        58001    57993       -8     
- Misses        976      981       +5     

Impacted Files	Coverage Δ
spec/support/http_helpers.rb	90.90% <0.00%> (-9.10%)	⬇️
spec/support/synchronization_helpers.rb	81.57% <0.00%> (-2.64%)	⬇️
lib/datadog/core/diagnostics/environment_logger.rb	98.42% <0.00%> (-1.58%)	⬇️
...adog/tracing/contrib/sidekiq/server_tracer_spec.rb	99.20% <0.00%> (-0.80%)	⬇️
...ng/contrib/active_support/cache/instrumentation.rb	87.58% <0.00%> (-0.09%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[marcotc]

Waiting on sorbet/sorbet#4119 to be addressed. There's already a PR up for it, currently under review.

[ivoanjo]

Waiting on sorbet/sorbet#4119 to be addressed. There's already a PR up for it, currently under review.

Note that upstream seems very slow on this issue, so I don't think we can expect a fix anytime soon 😰

[ivoanjo]

Note that there's two changes being made here: enabling it on arm64 for macOS and for Linux. Afaik it doesn't work for Linux /yet/, so I think we should leave that part of the check for now.

[lloeki]

but M1 Mac with arm64-darwin21 architecture is doing fin

aarch64-linux is not! This can't be merged as is.

This PR from @marco is the only sane way to do it: #2421

[lloeki]

With the ruby platform check, the appraisal lockfile would always end up with difference

Important note: the locks that needs to be commited to work under aarch64-linux MUST NOT have sorbet, as this prevents bundle lock --lockfile=ruby_x.y.z_foo.lock --add-platform aarch64-linux from working.

Ideally the appraisal lockfiles shoud all have PLAFTORMS with both x86_64-linux and aarch64-linux but not ruby.

As I said to @marcotc the other day, conditionals in Gemfiles are bad as they are evaluated on the currently run RUBY_PLATFORM + RUBY_VERSION, which has terrible effects on the descriptiveness of the DSL preventing proper cross platform + cross version usage of bundler.

I have a tentative change locally to address the Gemfile conditionals, making it entirely descriptive but didn't push a PR yet because I have a path issue to deal with when using the appraisal command.

[lloeki]

Note that upstream seems very slow on this issue, so I don't think we can expect a fix anytime soon

It's been working on a fork for something like a year. It's clearly just not their priority.

[ivoanjo]

Important note: the locks that needs to be commited to work under aarch64-linux MUST NOT have sorbet, as this prevents bundle lock --lockfile=ruby_x.y.z_foo.lock --add-platform aarch64-linux from working.

👍 thanks for the reminder. To be honest we don't need sorbet for the appraisal steps, so even without #2421 I think we could fix this by for instance tweaking the condition so that it doesn't trigger when the Gemfile is being evaluated for the appraisals.

But #2421 does seem to be a less hacky way of doing it, so that's probably the way to go.