Repository navigation
🐛 fix(type): pass the latest ty and close Any gaps - #4075
Merged
Merged
Conversation
gaborbernat
added
the
type:internal
gaborbernat
changed the title
gaborbernat
force-pushed
the
fix-ty-name-collision
branch
2 times, most recently
from
September 19, 2026 00:17
bd8fdf3 to
4040136
Compare
gaborbernat
changed the title
gaborbernat
force-pushed
the
fix-ty-name-collision
branch
from
September 19, 2026 00:31
4040136 to
28d2f4a
Compare
A ty release after main's last green run turned the type jobs red on an unchanged commit. Inside the ini loader package the bare name `replace` means both the substitution function imported there and the package's own `replace` submodule, because importing a submodule binds its name onto the parent package. ty now resolves the name to the submodule and rejects the calls. The function is imported under an unambiguous alias instead, which keeps the documented `tox.config.loader.ini.replace` import path intact. ty also reports instantiating an abstract class, which one test does on purpose, so that suppression gains the new code and every remaining suppression now says why it has to stay. Running ty with its opt-in rules showed how much the default run lets through. About sixty sites annotated a config lookup with a type the checker could not verify, since `conf[...]` returns Any; they now go through `ConfigSet.get`, and the new `ConfigSet.get_optional` covers keys that may be unset. Each converted key was checked against its registration: declared type, default and post-processing all produce the type asked for, so the runtime check cannot fire for tox's own configuration. The pass surfaced annotations that were wrong: `min_version` was declared Version but is None when unset, and dependency groups were typed as a list of strings or one include table rather than a list of either. Guards that could never be false are gone where the code right next to them already dereferences the same value, and the release task's ignores hid a Version compared against tag names and a Commit passed as a ref string. Every overriding method in src and tests carries `@override`, from typing on 3.12+ and typing_extensions before that. The opt-in rules that now come out clean are enabled so they stay clean.
gaborbernat
force-pushed
the
fix-ty-name-collision
branch
from
September 19, 2026 00:53
28d2f4a to
43d7636
Compare
gaborbernat
enabled auto-merge (squash)
September 19, 2026 00:54
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The type check jobs went red on
mainwith no code change: the scheduled run passed on 17 September and failed a day later on the same commit, because CI picked upty0.0.82. Every open PR now inherits that failure, #4074 included. 🔴The new release reports two things. Inside the ini loader package, the bare name
replacemeans both the substitution function and the package's ownreplacesubmodule, since importing a submodule binds its name onto the parent package. The function is now imported under an unambiguous alias, and thetox.config.loader.ini.replaceimport path stays as it was.tyalso rejects instantiating an abstract class. One test does that on purpose to prove the runtime rejects the call too, so its suppression gains the new diagnostic code, and every suppression that remains now states in place why it has to stay.I then ran
tywith its opt-in rules to see what the default run lets through, and fixed what had a real fix. Around sixty places read a configuration value asAnyand annotated it with a type nothing verified; they now go through the type-checked accessor added in 4.59, and a newConfigSet.get_optionalcovers values that may be unset. That exposed annotations that were wrong:min_versionwas declared as a version but isNonewhen unset, and dependency groups were typed as a list of strings or a single include table instead of a list of either. Every method that overrides a parent now carries@override. The release task's ignores hid a version object compared against tag names and a commit passed where a ref string is expected. ✅This is meant to change no behaviour. For each converted configuration key I checked its registration (declared type, default, post-processing) and all of them produce the type the accessor asks for, so its
TypeErrorcannot fire for tox's own configuration. An independent review compared 54 command invocations between this branch andmainacrosstox.iniandtox.tomlprojects, with and without tox-uv, tox-gh and tox-gh-actions, and found identical output; the tox-uv test suite fails the same tests on both. The opt-in rules that now come out clean are enabled so they stay that way. What remains needs separate decisions: command line options arrive as an untyped argparse namespace, and the platform-specific findings come from checking all platforms at once.