Repository navigation
Green main: fix the scaffold settings fixture, make two tests Windows-portable - #281
Merged
Merged
Conversation
The fake settings store used make-hasheq with string keys. Racket interns string literals per module, so the production key from resolve-rollout-bucket! never matched the test's literal: the draw persisted into a slot the test's read-back could not see, and the sticky assertion only held because the direct writer write landed in a third slot. Every platform failed 1/10017; main has been red since #271. Switch the fixture to make-hash and rewrite the sticky block to assert what the documented semantics promise: the draw is recovered, and a valid value written directly through the adapter is what the next run recovers. Drop the diagnostic eprintf that leaked in with the block.
project-report returns path->string output, so the AgentDemoApp expectation needs host separators, not hardcoded forward slashes; and Unix permission bits do not exist on NTFS, where everything reports 0o777, so the 0o755 maintainer-script assertion only runs on POSIX. Together with the previous commit the full suite passes on a Chinese-locale Windows 11 machine: 10011/10011.
The release-gate fixture wrote the snapshot to a cwd-relative path, and raco test runs each file with its own directory as cwd, so every full suite run left rivet-schema.json behind in rivet/tests/ — invisible on CI (untracked) but littering every developer checkout. The project-root path is also what the doctor's own advice tells downstreams to commit, so the green-path fixture now matches reality; the temp project is deleted in the existing dynamic-wind cleanup.
…faced Every sibling Windows project (the example and scaffold hosts, chart, device smoke) already passes /utf-8; only the static library the hosts link was missing it. On a CJK-locale machine MSVC then reads the UTF-8 sources in the system code page and drowns the build in C4819 (18 on a plain fulcrum raco rivet build), with real data-loss risk for any non-ASCII literal. With /utf-8 the remaining warnings are two C4244s in <xutility> instantiations and a C4456 pair in device_session_host.cpp where the expiry branch redeclared saw_acceptance/saw_close over the enclosing scope — renamed to expired_acceptance/expired_close, semantics unchanged.
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
1 participant
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.
Summary
The full suite has failed 1/10017 on every platform since #271 landed, so main went red on the merge commit itself. Reproduced and fixed locally on a Chinese-locale Windows 11 machine (Racket CS 9.0, 10011/10011 after both commits).
1. The red-main failure:
updater-scaffold.rktfixture (make-hasheq+ string keys)The fake settings store used
make-hasheq. Racket interns string literals per module, so the"update.rollout-bucket"literal insiderivet/distribution/updater-scaffold.rktand the one inside the test areeq?-distinct objects:resolve-rollout-bucket!persists the draw under its own literal; the test's read-back with its own literal misses (stored=#f) — this is the CI failure atupdater-scaffold.rkt:176.resolvereturnsfirst-bucketafter the direct"42"write) only ever passed because the direct write landed in a third, invisible slot. Under correctequal?semantics a valid"42"write is the recovered value, so the block's final expectation contradicted the documented semantics and had to change.Fixture is now
make-hash; the sticky block asserts the documented behavior (draw persisted, draw recovered, direct valid write recovered). The diagnosticeprintfthat leaked in is gone.2. Windows-host portability (fails on any Windows dev machine, invisible on CI's Linux)
inspect.rkt:project-reportreturnspath->stringoutput; the expectation now builds the path so separators follow the host instead of hardcoding/.linux-native-installer.rkt: Unix permission bits do not exist on NTFS (everything reports 0o777); the 0o755 maintainer-script assertion now runs on POSIX only.Test plan
raco test rivet/tests/updater-scaffold.rkt— 47 passedraco test rivet/tests/— 10011 passed, 0 failures (Windows 11, GBK code page)