Update gems and their RBI files #1246
Conversation
|
It seems that the new version of Sorbet reverts the attr_* runtime checks which is causing the test failures (we fixed them on #1227). We'll need to revert those tests back until the attr_* runtime checks are shipped again. |
dirceu
force-pushed
the
update-gems-20221025
branch
from
875eae8
to
fe0242b
Compare
|
@vinistock I reverted that and it caused failures in some of the matrix's pipelines; I removed the revert commit and other ones are failing. I'm a bit lost here as to why the |
|
The reason is probably because we don't commit a One option we can take is check the Sorbet version in the test and then change the generated RBI based on it. Something like expected_foo_signature = if sorbet_version < 123123 && sorbet_version > 123123
"def foo=(arg0); end"
else
<<~RUBY
sig { ... }
def foo=(foo); end
RUBY
end |
1. Latest Sorbet reverted a runtime check, so we should revert the "foo" signature we changed for that version too. More info: #1246 (comment) 2. Because the extra gemfiles don't lock Sorbet version, it always uses the latest one. But the tests running with Gemfile runs against the old Sorbet. So we get inconsistant CI result on the same test. 3. To make all tests pass, we need to do the 2 things at the same time: 1. Upgrade Sorbet 2. Revert the signature
1. Latest Sorbet reverted a runtime check, so we should revert the "foo" signature we changed for that version too. More info: #1246 (comment) 2. Because the extra gemfiles don't lock Sorbet version, it always uses the latest one. But the tests running with Gemfile runs against the old Sorbet. So we get inconsistant CI result on the same test. 3. To make all tests pass, we need to do the 2 things at the same time: 1. Upgrade Sorbet 2. Revert the signature
|
I solved the issue in #1247 so this PR should be passing after rebasing/merging |
Closes #1242
Closes #1243
Closes #1244
Closes #1245