fix: Reviewer Disposition dark-mode contrast + registry manifest hash refresh - #169
Conversation
…ash refresh - Reviewer Disposition select/input had no dark: classes (same pattern as the Acceptance Criteria fix in v3.4.1), unreadable in dark mode. Scanned the rest of the page for the same pattern: all other bare select/input elements are sr-only toggle/radio inputs styled via sibling elements, not affected. - reproducibility/benchmark_v0.5.1/MANIFEST.json's pinned parameter_registry hash drifted because the v3.4.1 release.py bump touched the registry YAML's internal `version:` field (a schema-version label, not the package version, but release.py's string-replace can't tell them apart). Confirmed via diff that the only change to the registry file was that one version string (no parameter/weight changes), so refreshing the pinned hash is safe and matches the established precedent (commit 999b472). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d586342657
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "parameter_registry": { | ||
| "path": "src/factorforge/registry/current_parameter_registry.yaml", | ||
| "sha256": "1cc163faa5a5be118b0d221a63a93a1241f1f9d82d85ab6d4c627c8988b0542a" | ||
| "sha256": "64fd2f9de792b215ce257c8300e01cb1d86765f0a726740a8893aee6d8dff5e6" |
There was a problem hiding this comment.
When reproducing benchmark_v0.5.1, this checksum now verifies the current src/factorforge/registry/current_parameter_registry.yaml blob, whose registry version is 3.4.1, while the evidence pack itself still advertises FactorForge/registry 3.2.0 in reproducibility/benchmark_v0.5.1/README.md:13-15. Because the current registry has since changed benchmark-relevant defaults such as the N. benthamiana GC band and codon reference, the manifest can pass while proving the wrong registry for the v3.2.0 results; keep this input pinned to the actual frozen v3.2.0 registry artifact instead of refreshing it to HEAD.
Useful? React with 👍 / 👎.
Full-page scan for color utilities not covered by the global .dark override rules in style.css (which already handle bg-white, bg-slate-50/100, text-slate-900/800/700/600/500/400, and border-slate-200/100 automatically). Five small gaps found, all missing dark: variants entirely (not text-invisible bugs, but visible light-mode-colored elements on a dark page): - Clear Input / Copy buttons: hover background had no dark: variant (copied the working pattern already used elsewhere on the page) - File upload dropzone: dashed border color - "Standard Expression" checkbox border - Hidden DNA-type badge background (same pattern as the adjacent "N. benthamiana only" / "Experimental" badges, which already had it) - Footer separator dot Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
1 participant
Summary
dark:-classes pattern as the v3.4.1 fix, this time in the "Reviewer Disposition" select/input (web/index.html). Scanned the rest of the page for the same pattern — everything else withoutdark:classes is asr-onlytoggle/radio input styled via sibling elements, not user-visible text.reproducibility/benchmark_v0.5.1/MANIFEST.json's pinnedparameter_registryhash drifted because the v3.4.1 release.py bump touched the registry YAML's internalversion:field. Confirmed viagit diffthat the only change to the registry file itself was that one version string — no parameter/weight changes — so refreshing the pinned hash is safe and matches the established precedent (commit 999b472, same fix pattern used before).Test plan
pytest -q— 766 passed, 2 skipped (was 1 failing on manifest drift before this fix)Follow-up (not in this PR)
scripts/release.py's string-replace bump touchessrc/factorforge/registry/current_parameter_registry.yaml'sversion:field whenever it happens to equal the old package version string, even though that field is an independent parameter-registry schema version, not the package version. This is a blind spot worth fixing in release.py itself so future releases don't silently re-break this manifest pin.🤖 Generated with Claude Code