Adds doc snippet compile validation; fixes v1 API rot in agent docs#5480
Merged
Conversation
![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: 2f7956e8f6
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".
tig
added a commit
that referenced
this pull request
Jun 10, 2026
Codex review on #5480: the validator only failed on Severity==Error, so a doc snippet using a v1-compatible [Obsolete] API (e.g. the legacy static Application.Init) compiled "clean" — letting the exact v1 rot this workflow prevents slip back in. Test-first: added testdata/obsolete-api.md (an obsolete-API snippet, not marked WRONG, excluded from the validated set) and a CI negative-test step asserting the validator rejects it. With the old behavior the fixture passed (red). Fix: elevate CS0618/CS0612 to errors via specificDiagnosticOptions, and drop the blanket `#pragma warning disable` from the wrappers (it would otherwise suppress those very warnings; harness noise stays warnings, which never fail). The negative test now passes (red→green). This immediately caught real rot in the cookbook: two snippets used the now-obsolete TextView (superseded by gui-cs/Editor's EditorView). They were layout examples, so the incidental editor is now a plain content View with a note pointing at EditorView. Real docs: 29 compiled, 1 skipped, 0 failed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Roslyn-based validator that compiles every fenced C# block in the agent docs (ai-v2-primer.md, build-app.md, common-patterns.md) against the built Terminal.Gui assembly, so example rot fails CI instead of misleading readers. Statement fragments are wrapped in a Runnable<string?> harness; blocks marked WRONG / `snippet: ignore` are skipped. Obsolete-API use (CS0618/CS0612) is treated as a failure, not a warning, so v1 rot like the legacy static Application.Init cannot pass as "compiled". testdata/obsolete-api.md + a CI negative-test step assert the validator rejects obsolete APIs. The check caught real rot fixed here: two cookbook snippets used the obsolete TextView (now a plain content View, pointing at gui-cs/Editor's EditorView). Result: 29 compiled, 1 skipped, 0 failed; negative test green. Rebased onto develop after #5478 merged (squash), keeping only this PR's files; no overlap with the merged agent-docs lint changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
tig
force-pushed
the
feature/doc-snippet-validation
branch
from
531b87d to
34668cf
Compare
The workflow built Terminal.Gui without fetch-depth:0, so GitVersion.MsBuild
failed on the shallow clone ("Repository is a shallow clone"). Match the
build-validation workflow's checkout (fetch-depth: 0).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
tig
added a commit
that referenced
this pull request
Jun 12, 2026
* Fixes #5434. Track adornment redraw deltas Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Clean up adornment thickness-delta helpers Fix indentation of the Margin.ThicknessChanged handler body and factor the duplicated viewport-size formula out of GetViewportForAdornmentThickness and GetViewportFrameForAdornmentThickness into a shared GetViewportSizeForThickness helper. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Bump TextMateSharp and TextMateSharp.Grammars Bumps TextMateSharp from 2.0.3 to 2.0.4 Bumps TextMateSharp.Grammars from 2.0.3 to 2.0.4 --- updated-dependencies: - dependency-name: TextMateSharp dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch - dependency-name: TextMateSharp.Grammars dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> * Regenerate docfx Mandelbrot.gif with rendered sixel (#5471) * Bump Spectre.Console from 0.55.2 to 0.56.0 (#5475) --- updated-dependencies: - dependency-name: Spectre.Console dependency-version: 0.56.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Fixes #5360. Cull fully occluded overlapped opaque subviews during draw DrawSubViews() now skips drawing an Overlapped sibling when it is entirely covered by the higher-Z opaque peers already drawn. Such a sibling produces no visible output — every cell it would draw, including its own RenderLineCanvas, is clipped away by the clip "holes" those peers punched in DoDrawComplete — so skipping its Draw is output-neutral. The culled view's NeedsDraw is cleared (mirroring a drawn-but-fully-clipped pass) so it does not keep the SuperView perpetually dirty. Culling is intentionally conservative — it only applies to a candidate that is itself Overlapped, opaque (content/Border/Padding not Transparent), without a Margin shadow, and NOT SuperViewRendersLineCanvas. The last guard is what keeps Tabs correct: tab pages render their borders (and headers) through the SuperView's LineCanvas painters'-algorithm composition, so they must never be culled or their header line art would be dropped. The whole check is gated on the presence of an Overlapped sibling, keeping the common non-overlapping draw path zero-overhead. Opaque coverage is the region inside a view's (transparent-by-default) Margin — the Border frame — which matches the area DoDrawComplete excludes from the clip. Adds OcclusionCullingTests covering: fully-occluded cull, NeedsDraw clearing, partial coverage, transparent occluder, transparent candidate, shadowed candidate, SuperViewRendersLineCanvas candidate, non-Overlapped candidate, the no-Overlapped-siblings gate, output-neutrality, and a Tabs regression that inactive pages are never culled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Address review: don't cull TransparentMouse candidates The cull path skips Draw()/DoDrawComplete(), which repopulates a view's CachedDrawnRegion for ViewportSettingsFlags.TransparentMouse layers. That region is invalidated by SetNeedsDraw(), so culling an opaque TransparentMouse view left it with a null cache and dropped it from mouse hit-testing (GetViewsUnderLocation blanket-removes TransparentMouse views with a null cache) — a state divergence from the drawn-but-clipped path the cull is meant to mirror. Add ParticipatesInTransparentMouseHitTesting() and exclude such candidates: the view's own ViewportSettings, plus any adornment that is TransparentMouse with non-empty Thickness (adornment caches survive SetNeedsDraw(), but the first draw must still populate a thick transparent-mouse adornment's cache). Margin is TransparentMouse by default, so this only excludes views with an actual Margin thickness — the common empty-margin overlapped view is still culled. Adds regression tests: TransparentMouseCandidate_IsNotCulled_AndKeepsHitRegion (fails without the guard — CachedDrawnRegion is null) and TransparentMouseMarginCandidate_IsNotCulled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Fixes stale/contradictory AI agent instructions; adds agent docs lint and tuirec verification guidance (#5478) * Fix inconsistencies in AI agent instruction files - Local function naming: camelCase -> PascalCase in REFRESH.md and copilot-instructions.md (matches .editorconfig local_functions_rule, AGENTS.md, and event-patterns.md) - Replace stale Tests/UnitTests references with current project names (UnitTestsParallelizable / UnitTests.NonParallelizable) across AGENTS.md, CONTRIBUTING.md, .aider.md, .cursorrules, .windsurfrules, copilot-instructions.md, and .claude workflows/tasks - Replace deprecated --filter "FullyQualifiedName~" syntax with xUnit v3 MTP --filter-method/--filter-class in copilot-instructions.md - Remove machine-local path (D:\s\...) from AGENTS.md planning section - build-app.md: use Accepted (post-event) for fire-and-forget handlers per event-patterns.md; show Accepting only for cancellation; fix v1-style new Button ("OK") to v2 object initializer - build-test-workflow.md: .NET SDK 8.0 -> 10.0.100 per global.json Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Add agent docs lint, tuirec verification guidance, promote local memories - Add Scripts/lint-agent-docs.ps1 + lint-agent-docs.yml CI workflow: fails when known rot patterns reappear in agent instruction files (stale test project names, machine-local paths, camelCase local-function guidance, deprecated FullyQualifiedName~ filter syntax, SDK version drift vs global.json). The lint immediately caught a stale .NET SDK 8.0 claim in CONTRIBUTING.md, now fixed. - Wire tuirec into agent entry points (CLAUDE.md, AGENTS.md, build-app.md): agents can verify TUI behavior by recording with tuirec and reading the asciinema .cast output back as text, per Scripts/tuirec/README.md. - Promote durable guidance from machine-specific .claude/projects/ memory files into shared rules (.claude/rules/logging-tracing.md and fragile-areas.md); untrack .claude/projects/ and gitignore it (the directory name encoded a local user path). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Document Windows ConPTY sixel limitation in tuirec README Discovered while verifying the About Box fire animation: ConPTY strips sixel DCS and the DA1 sixel handshake, so sixel content cannot be captured in tuirec recordings on Windows (apps detect Sixel support: False). Added to the troubleshooting table and validation checklist so the next agent does not burn recordings rediscovering it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * tuirec README: require measuring grid-anchored sixels, not eyeballing Agents (including me) recurrently verify the wrong invariant when checking sixel recordings: confirming the sixel appears, or that agg rendered it faithfully at the requested cursor cell, and calling it done. That misses size/position errors — notably the ~4% undersize from tuirec advertising a cell resolution that does not match agg's rendered font cell (tuirec #84). Adds a "Verifying Placement and Size (measure - don't eyeball)" section with the cell-calibration recipe (measure agg's real cell from a known grid reference; reconcile against the resolution the app used; confirm the rendered bbox covers the target region), a checklist item, a troubleshooting row for #84, and rewrites the workflow's "visual confirm" step to require measurement for grid-anchored content. Also states the general principle: verify the invariant the change was meant to satisfy, not a proxy. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Sync CONTRIBUTING.md TFM to net10.0; lint stale TFMs Codex review on #5478 flagged that build-test-workflow.md cites CONTRIBUTING.md as its source of truth, but CONTRIBUTING.md still described the project as net8.0 (line 28) even though the Required Tools section was updated to .NET 10 — so the declared source could revert the fix and mislead readers onto the wrong toolchain. Test-first: added Rule 6 to lint-agent-docs.ps1 that derives the expected target-framework moniker from global.json's SDK major and fails on any mismatched `net<major>.0` reference. It flagged CONTRIBUTING.md:28 (net8.0); fixed that to `C# 14 (net10.0)`. Lint now passes (32 files). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Adds doc snippet compile validation; fixes v1 API rot in agent docs (#5480) * Add doc snippet compile validation with obsolete-API rejection Roslyn-based validator that compiles every fenced C# block in the agent docs (ai-v2-primer.md, build-app.md, common-patterns.md) against the built Terminal.Gui assembly, so example rot fails CI instead of misleading readers. Statement fragments are wrapped in a Runnable<string?> harness; blocks marked WRONG / `snippet: ignore` are skipped. Obsolete-API use (CS0618/CS0612) is treated as a failure, not a warning, so v1 rot like the legacy static Application.Init cannot pass as "compiled". testdata/obsolete-api.md + a CI negative-test step assert the validator rejects obsolete APIs. The check caught real rot fixed here: two cookbook snippets used the obsolete TextView (now a plain content View, pointing at gui-cs/Editor's EditorView). Result: 29 compiled, 1 skipped, 0 failed; negative test green. Rebased onto develop after #5478 merged (squash), keeping only this PR's files; no overlap with the merged agent-docs lint changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Fix Validate Doc Snippets CI: full history for GitVersion The workflow built Terminal.Gui without fetch-depth:0, so GitVersion.MsBuild failed on the shallow clone ("Repository is a shallow clone"). Match the build-validation workflow's checkout (fetch-depth: 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Notify gui-cs/Editor on NuGet publish (#5483) Dispatches `terminal-gui-published` (payload: the published SemVer) to gui-cs/Editor after each publish. Editor's bump-terminal-gui.yml picks it up, pins the new version, validates with its full test suites, and republishes Editor against it - making Editor a continuous canary for TG API churn (see Editor specs/decisions.md DEC-010). No-op until an EDITOR_DISPATCH_TOKEN secret (PAT with repo scope on gui-cs/Editor) is configured; Editor also polls NuGet on a 6h schedule as a fallback, so nothing breaks without it. Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Fix publish.yml: secrets context is invalid in step if conditions The Editor-notify guard from #5483 used `if: secrets.X != ''`, which is not an available context for step ifs and invalidated the whole workflow (publish run 27422643812). Mirror the secret into job-level env and gate on that, same pattern as the TEMPLATE/clet dispatches in gui-cs/Editor. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
This was referenced Jun 22, 2026
tig
added a commit
that referenced
this pull request
Jun 22, 2026
* Release v2.4.5 (#5476) * Fixes #5434. Track adornment redraw deltas Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Clean up adornment thickness-delta helpers Fix indentation of the Margin.ThicknessChanged handler body and factor the duplicated viewport-size formula out of GetViewportForAdornmentThickness and GetViewportFrameForAdornmentThickness into a shared GetViewportSizeForThickness helper. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Bump TextMateSharp and TextMateSharp.Grammars Bumps TextMateSharp from 2.0.3 to 2.0.4 Bumps TextMateSharp.Grammars from 2.0.3 to 2.0.4 --- updated-dependencies: - dependency-name: TextMateSharp dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch - dependency-name: TextMateSharp.Grammars dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> * Regenerate docfx Mandelbrot.gif with rendered sixel (#5471) * Bump Spectre.Console from 0.55.2 to 0.56.0 (#5475) --- updated-dependencies: - dependency-name: Spectre.Console dependency-version: 0.56.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Release v2.4.6 (#5485) * Fixes #5434. Track adornment redraw deltas Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Clean up adornment thickness-delta helpers Fix indentation of the Margin.ThicknessChanged handler body and factor the duplicated viewport-size formula out of GetViewportForAdornmentThickness and GetViewportFrameForAdornmentThickness into a shared GetViewportSizeForThickness helper. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Bump TextMateSharp and TextMateSharp.Grammars Bumps TextMateSharp from 2.0.3 to 2.0.4 Bumps TextMateSharp.Grammars from 2.0.3 to 2.0.4 --- updated-dependencies: - dependency-name: TextMateSharp dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch - dependency-name: TextMateSharp.Grammars dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> * Regenerate docfx Mandelbrot.gif with rendered sixel (#5471) * Bump Spectre.Console from 0.55.2 to 0.56.0 (#5475) --- updated-dependencies: - dependency-name: Spectre.Console dependency-version: 0.56.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Fixes #5360. Cull fully occluded overlapped opaque subviews during draw DrawSubViews() now skips drawing an Overlapped sibling when it is entirely covered by the higher-Z opaque peers already drawn. Such a sibling produces no visible output — every cell it would draw, including its own RenderLineCanvas, is clipped away by the clip "holes" those peers punched in DoDrawComplete — so skipping its Draw is output-neutral. The culled view's NeedsDraw is cleared (mirroring a drawn-but-fully-clipped pass) so it does not keep the SuperView perpetually dirty. Culling is intentionally conservative — it only applies to a candidate that is itself Overlapped, opaque (content/Border/Padding not Transparent), without a Margin shadow, and NOT SuperViewRendersLineCanvas. The last guard is what keeps Tabs correct: tab pages render their borders (and headers) through the SuperView's LineCanvas painters'-algorithm composition, so they must never be culled or their header line art would be dropped. The whole check is gated on the presence of an Overlapped sibling, keeping the common non-overlapping draw path zero-overhead. Opaque coverage is the region inside a view's (transparent-by-default) Margin — the Border frame — which matches the area DoDrawComplete excludes from the clip. Adds OcclusionCullingTests covering: fully-occluded cull, NeedsDraw clearing, partial coverage, transparent occluder, transparent candidate, shadowed candidate, SuperViewRendersLineCanvas candidate, non-Overlapped candidate, the no-Overlapped-siblings gate, output-neutrality, and a Tabs regression that inactive pages are never culled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Address review: don't cull TransparentMouse candidates The cull path skips Draw()/DoDrawComplete(), which repopulates a view's CachedDrawnRegion for ViewportSettingsFlags.TransparentMouse layers. That region is invalidated by SetNeedsDraw(), so culling an opaque TransparentMouse view left it with a null cache and dropped it from mouse hit-testing (GetViewsUnderLocation blanket-removes TransparentMouse views with a null cache) — a state divergence from the drawn-but-clipped path the cull is meant to mirror. Add ParticipatesInTransparentMouseHitTesting() and exclude such candidates: the view's own ViewportSettings, plus any adornment that is TransparentMouse with non-empty Thickness (adornment caches survive SetNeedsDraw(), but the first draw must still populate a thick transparent-mouse adornment's cache). Margin is TransparentMouse by default, so this only excludes views with an actual Margin thickness — the common empty-margin overlapped view is still culled. Adds regression tests: TransparentMouseCandidate_IsNotCulled_AndKeepsHitRegion (fails without the guard — CachedDrawnRegion is null) and TransparentMouseMarginCandidate_IsNotCulled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Fixes stale/contradictory AI agent instructions; adds agent docs lint and tuirec verification guidance (#5478) * Fix inconsistencies in AI agent instruction files - Local function naming: camelCase -> PascalCase in REFRESH.md and copilot-instructions.md (matches .editorconfig local_functions_rule, AGENTS.md, and event-patterns.md) - Replace stale Tests/UnitTests references with current project names (UnitTestsParallelizable / UnitTests.NonParallelizable) across AGENTS.md, CONTRIBUTING.md, .aider.md, .cursorrules, .windsurfrules, copilot-instructions.md, and .claude workflows/tasks - Replace deprecated --filter "FullyQualifiedName~" syntax with xUnit v3 MTP --filter-method/--filter-class in copilot-instructions.md - Remove machine-local path (D:\s\...) from AGENTS.md planning section - build-app.md: use Accepted (post-event) for fire-and-forget handlers per event-patterns.md; show Accepting only for cancellation; fix v1-style new Button ("OK") to v2 object initializer - build-test-workflow.md: .NET SDK 8.0 -> 10.0.100 per global.json Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Add agent docs lint, tuirec verification guidance, promote local memories - Add Scripts/lint-agent-docs.ps1 + lint-agent-docs.yml CI workflow: fails when known rot patterns reappear in agent instruction files (stale test project names, machine-local paths, camelCase local-function guidance, deprecated FullyQualifiedName~ filter syntax, SDK version drift vs global.json). The lint immediately caught a stale .NET SDK 8.0 claim in CONTRIBUTING.md, now fixed. - Wire tuirec into agent entry points (CLAUDE.md, AGENTS.md, build-app.md): agents can verify TUI behavior by recording with tuirec and reading the asciinema .cast output back as text, per Scripts/tuirec/README.md. - Promote durable guidance from machine-specific .claude/projects/ memory files into shared rules (.claude/rules/logging-tracing.md and fragile-areas.md); untrack .claude/projects/ and gitignore it (the directory name encoded a local user path). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Document Windows ConPTY sixel limitation in tuirec README Discovered while verifying the About Box fire animation: ConPTY strips sixel DCS and the DA1 sixel handshake, so sixel content cannot be captured in tuirec recordings on Windows (apps detect Sixel support: False). Added to the troubleshooting table and validation checklist so the next agent does not burn recordings rediscovering it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * tuirec README: require measuring grid-anchored sixels, not eyeballing Agents (including me) recurrently verify the wrong invariant when checking sixel recordings: confirming the sixel appears, or that agg rendered it faithfully at the requested cursor cell, and calling it done. That misses size/position errors — notably the ~4% undersize from tuirec advertising a cell resolution that does not match agg's rendered font cell (tuirec #84). Adds a "Verifying Placement and Size (measure - don't eyeball)" section with the cell-calibration recipe (measure agg's real cell from a known grid reference; reconcile against the resolution the app used; confirm the rendered bbox covers the target region), a checklist item, a troubleshooting row for #84, and rewrites the workflow's "visual confirm" step to require measurement for grid-anchored content. Also states the general principle: verify the invariant the change was meant to satisfy, not a proxy. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Sync CONTRIBUTING.md TFM to net10.0; lint stale TFMs Codex review on #5478 flagged that build-test-workflow.md cites CONTRIBUTING.md as its source of truth, but CONTRIBUTING.md still described the project as net8.0 (line 28) even though the Required Tools section was updated to .NET 10 — so the declared source could revert the fix and mislead readers onto the wrong toolchain. Test-first: added Rule 6 to lint-agent-docs.ps1 that derives the expected target-framework moniker from global.json's SDK major and fails on any mismatched `net<major>.0` reference. It flagged CONTRIBUTING.md:28 (net8.0); fixed that to `C# 14 (net10.0)`. Lint now passes (32 files). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Adds doc snippet compile validation; fixes v1 API rot in agent docs (#5480) * Add doc snippet compile validation with obsolete-API rejection Roslyn-based validator that compiles every fenced C# block in the agent docs (ai-v2-primer.md, build-app.md, common-patterns.md) against the built Terminal.Gui assembly, so example rot fails CI instead of misleading readers. Statement fragments are wrapped in a Runnable<string?> harness; blocks marked WRONG / `snippet: ignore` are skipped. Obsolete-API use (CS0618/CS0612) is treated as a failure, not a warning, so v1 rot like the legacy static Application.Init cannot pass as "compiled". testdata/obsolete-api.md + a CI negative-test step assert the validator rejects obsolete APIs. The check caught real rot fixed here: two cookbook snippets used the obsolete TextView (now a plain content View, pointing at gui-cs/Editor's EditorView). Result: 29 compiled, 1 skipped, 0 failed; negative test green. Rebased onto develop after #5478 merged (squash), keeping only this PR's files; no overlap with the merged agent-docs lint changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Fix Validate Doc Snippets CI: full history for GitVersion The workflow built Terminal.Gui without fetch-depth:0, so GitVersion.MsBuild failed on the shallow clone ("Repository is a shallow clone"). Match the build-validation workflow's checkout (fetch-depth: 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Notify gui-cs/Editor on NuGet publish (#5483) Dispatches `terminal-gui-published` (payload: the published SemVer) to gui-cs/Editor after each publish. Editor's bump-terminal-gui.yml picks it up, pins the new version, validates with its full test suites, and republishes Editor against it - making Editor a continuous canary for TG API churn (see Editor specs/decisions.md DEC-010). No-op until an EDITOR_DISPATCH_TOKEN secret (PAT with repo scope on gui-cs/Editor) is configured; Editor also polls NuGet on a 6h schedule as a fallback, so nothing breaks without it. Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Fix publish.yml: secrets context is invalid in step if conditions The Editor-notify guard from #5483 used `if: secrets.X != ''`, which is not an available context for step ifs and invalidated the whole workflow (publish run 27422643812). Mirror the secret into job-level env and gate on that, same pattern as the TEMPLATE/clet dispatches in gui-cs/Editor. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Mandelbrot: fill display with image view + regenerate GIF with Kitty graphics (#5511) * Regenerate Mandelbrot.gif with Kitty graphics rendering The previous capture forced the Sixel raster protocol because agg could not render Kitty graphics. tuirec v0.9.0 now defaults to Kitty graphics and its pinned agg (v1.11.0-sixel, built on a Kitty-capable avt) renders them, so the Mandelbrot scenario records through its preferred Kitty path. Recorded with the released tuirec + auto-downloaded agg. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Mandelbrot: fill the display with the image view; richer zoom/pan demo Anchor the MandelbrotImageView at (0,0) and size it Dim.Fill() so the fractal fills the whole display region instead of a small centered box, and move the status label to the bottom edge. Re-record the docfx GIF with a smoother tour: a multi-step zoom into the seahorse valley, a pan across the seahorse filaments, and a zoom back out — demonstrating the smooth zooming and panning the larger image view makes possible. * Scripts/tuirec: document Kitty graphics as the default raster path tuirec v0.9.0 advertises a Kitty identity by default, so raster apps (Mandelbrot, Images) now record via Kitty graphics, which agg v1.11.0-sixel renders. Add a 'Raster graphics' section covering protocol selection, how to confirm which protocol the cast captured, that the #84 cell-size mismatch is sixel-only, and a smooth zoom/pan recording tip. Update the validation checklist and troubleshooting for Kitty + sixel. * Scripts/tuirec: add exact one-shot recipe for Mandelbrot.gif Capture the precise keystroke tour, flags, and the design rationale (seahorse-valley target coords, center-anchored keyboard zoom, the span ~1.0 / 80-iteration black-out limit, fill layout) so a future agent can reproduce the hero GIF in one shot instead of re-deriving it. * Scripts/tuirec: cross-platform commands + runnable verification Address feedback from an agent that ran the guide on Linux: - Add bash equivalents alongside PowerShell for the raster recipes, validation greps, and error check (the raster recipes are Linux/macOS only, but every snippet was PowerShell). Make the Mandelbrot recipe bash-primary. - Add a fresh-container bootstrap (dotnet SDK matching global.json, Go install, PATH for ~/.dotnet and GOPATH/bin) and fix the Windows-only GOPATH PATH note. - Give a python3/Pillow one-liner to extract a mid-zoom frame, plus the expected readout (Center X ~ -0.74, Y ~ 0.105, Span ~ 1.0) so the spot-check measures a landmark instead of a vibe. - Soften exact payload counts to robust invariants (thousands of Kitty, zero sixel, ~0.9 MB) since the count drifts with timing/font-size. - Note the '#84 font-size adjust' log line is expected and auto-corrected in v0.9.0, harmless for the Kitty path. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
tig
added a commit
that referenced
this pull request
Jun 23, 2026
* Release v2.4.5 (#5476) * Fixes #5434. Track adornment redraw deltas Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Clean up adornment thickness-delta helpers Fix indentation of the Margin.ThicknessChanged handler body and factor the duplicated viewport-size formula out of GetViewportForAdornmentThickness and GetViewportFrameForAdornmentThickness into a shared GetViewportSizeForThickness helper. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Bump TextMateSharp and TextMateSharp.Grammars Bumps TextMateSharp from 2.0.3 to 2.0.4 Bumps TextMateSharp.Grammars from 2.0.3 to 2.0.4 --- updated-dependencies: - dependency-name: TextMateSharp dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch - dependency-name: TextMateSharp.Grammars dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> * Regenerate docfx Mandelbrot.gif with rendered sixel (#5471) * Bump Spectre.Console from 0.55.2 to 0.56.0 (#5475) --- updated-dependencies: - dependency-name: Spectre.Console dependency-version: 0.56.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Release v2.4.6 (#5485) * Fixes #5434. Track adornment redraw deltas Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Clean up adornment thickness-delta helpers Fix indentation of the Margin.ThicknessChanged handler body and factor the duplicated viewport-size formula out of GetViewportForAdornmentThickness and GetViewportFrameForAdornmentThickness into a shared GetViewportSizeForThickness helper. No behavior change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Bump TextMateSharp and TextMateSharp.Grammars Bumps TextMateSharp from 2.0.3 to 2.0.4 Bumps TextMateSharp.Grammars from 2.0.3 to 2.0.4 --- updated-dependencies: - dependency-name: TextMateSharp dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch - dependency-name: TextMateSharp.Grammars dependency-version: 2.0.4 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> * Regenerate docfx Mandelbrot.gif with rendered sixel (#5471) * Bump Spectre.Console from 0.55.2 to 0.56.0 (#5475) --- updated-dependencies: - dependency-name: Spectre.Console dependency-version: 0.56.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Fixes #5360. Cull fully occluded overlapped opaque subviews during draw DrawSubViews() now skips drawing an Overlapped sibling when it is entirely covered by the higher-Z opaque peers already drawn. Such a sibling produces no visible output — every cell it would draw, including its own RenderLineCanvas, is clipped away by the clip "holes" those peers punched in DoDrawComplete — so skipping its Draw is output-neutral. The culled view's NeedsDraw is cleared (mirroring a drawn-but-fully-clipped pass) so it does not keep the SuperView perpetually dirty. Culling is intentionally conservative — it only applies to a candidate that is itself Overlapped, opaque (content/Border/Padding not Transparent), without a Margin shadow, and NOT SuperViewRendersLineCanvas. The last guard is what keeps Tabs correct: tab pages render their borders (and headers) through the SuperView's LineCanvas painters'-algorithm composition, so they must never be culled or their header line art would be dropped. The whole check is gated on the presence of an Overlapped sibling, keeping the common non-overlapping draw path zero-overhead. Opaque coverage is the region inside a view's (transparent-by-default) Margin — the Border frame — which matches the area DoDrawComplete excludes from the clip. Adds OcclusionCullingTests covering: fully-occluded cull, NeedsDraw clearing, partial coverage, transparent occluder, transparent candidate, shadowed candidate, SuperViewRendersLineCanvas candidate, non-Overlapped candidate, the no-Overlapped-siblings gate, output-neutrality, and a Tabs regression that inactive pages are never culled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Address review: don't cull TransparentMouse candidates The cull path skips Draw()/DoDrawComplete(), which repopulates a view's CachedDrawnRegion for ViewportSettingsFlags.TransparentMouse layers. That region is invalidated by SetNeedsDraw(), so culling an opaque TransparentMouse view left it with a null cache and dropped it from mouse hit-testing (GetViewsUnderLocation blanket-removes TransparentMouse views with a null cache) — a state divergence from the drawn-but-clipped path the cull is meant to mirror. Add ParticipatesInTransparentMouseHitTesting() and exclude such candidates: the view's own ViewportSettings, plus any adornment that is TransparentMouse with non-empty Thickness (adornment caches survive SetNeedsDraw(), but the first draw must still populate a thick transparent-mouse adornment's cache). Margin is TransparentMouse by default, so this only excludes views with an actual Margin thickness — the common empty-margin overlapped view is still culled. Adds regression tests: TransparentMouseCandidate_IsNotCulled_AndKeepsHitRegion (fails without the guard — CachedDrawnRegion is null) and TransparentMouseMarginCandidate_IsNotCulled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Fixes stale/contradictory AI agent instructions; adds agent docs lint and tuirec verification guidance (#5478) * Fix inconsistencies in AI agent instruction files - Local function naming: camelCase -> PascalCase in REFRESH.md and copilot-instructions.md (matches .editorconfig local_functions_rule, AGENTS.md, and event-patterns.md) - Replace stale Tests/UnitTests references with current project names (UnitTestsParallelizable / UnitTests.NonParallelizable) across AGENTS.md, CONTRIBUTING.md, .aider.md, .cursorrules, .windsurfrules, copilot-instructions.md, and .claude workflows/tasks - Replace deprecated --filter "FullyQualifiedName~" syntax with xUnit v3 MTP --filter-method/--filter-class in copilot-instructions.md - Remove machine-local path (D:\s\...) from AGENTS.md planning section - build-app.md: use Accepted (post-event) for fire-and-forget handlers per event-patterns.md; show Accepting only for cancellation; fix v1-style new Button ("OK") to v2 object initializer - build-test-workflow.md: .NET SDK 8.0 -> 10.0.100 per global.json Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Add agent docs lint, tuirec verification guidance, promote local memories - Add Scripts/lint-agent-docs.ps1 + lint-agent-docs.yml CI workflow: fails when known rot patterns reappear in agent instruction files (stale test project names, machine-local paths, camelCase local-function guidance, deprecated FullyQualifiedName~ filter syntax, SDK version drift vs global.json). The lint immediately caught a stale .NET SDK 8.0 claim in CONTRIBUTING.md, now fixed. - Wire tuirec into agent entry points (CLAUDE.md, AGENTS.md, build-app.md): agents can verify TUI behavior by recording with tuirec and reading the asciinema .cast output back as text, per Scripts/tuirec/README.md. - Promote durable guidance from machine-specific .claude/projects/ memory files into shared rules (.claude/rules/logging-tracing.md and fragile-areas.md); untrack .claude/projects/ and gitignore it (the directory name encoded a local user path). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Document Windows ConPTY sixel limitation in tuirec README Discovered while verifying the About Box fire animation: ConPTY strips sixel DCS and the DA1 sixel handshake, so sixel content cannot be captured in tuirec recordings on Windows (apps detect Sixel support: False). Added to the troubleshooting table and validation checklist so the next agent does not burn recordings rediscovering it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * tuirec README: require measuring grid-anchored sixels, not eyeballing Agents (including me) recurrently verify the wrong invariant when checking sixel recordings: confirming the sixel appears, or that agg rendered it faithfully at the requested cursor cell, and calling it done. That misses size/position errors — notably the ~4% undersize from tuirec advertising a cell resolution that does not match agg's rendered font cell (tuirec #84). Adds a "Verifying Placement and Size (measure - don't eyeball)" section with the cell-calibration recipe (measure agg's real cell from a known grid reference; reconcile against the resolution the app used; confirm the rendered bbox covers the target region), a checklist item, a troubleshooting row for #84, and rewrites the workflow's "visual confirm" step to require measurement for grid-anchored content. Also states the general principle: verify the invariant the change was meant to satisfy, not a proxy. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Sync CONTRIBUTING.md TFM to net10.0; lint stale TFMs Codex review on #5478 flagged that build-test-workflow.md cites CONTRIBUTING.md as its source of truth, but CONTRIBUTING.md still described the project as net8.0 (line 28) even though the Required Tools section was updated to .NET 10 — so the declared source could revert the fix and mislead readers onto the wrong toolchain. Test-first: added Rule 6 to lint-agent-docs.ps1 that derives the expected target-framework moniker from global.json's SDK major and fails on any mismatched `net<major>.0` reference. It flagged CONTRIBUTING.md:28 (net8.0); fixed that to `C# 14 (net10.0)`. Lint now passes (32 files). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Adds doc snippet compile validation; fixes v1 API rot in agent docs (#5480) * Add doc snippet compile validation with obsolete-API rejection Roslyn-based validator that compiles every fenced C# block in the agent docs (ai-v2-primer.md, build-app.md, common-patterns.md) against the built Terminal.Gui assembly, so example rot fails CI instead of misleading readers. Statement fragments are wrapped in a Runnable<string?> harness; blocks marked WRONG / `snippet: ignore` are skipped. Obsolete-API use (CS0618/CS0612) is treated as a failure, not a warning, so v1 rot like the legacy static Application.Init cannot pass as "compiled". testdata/obsolete-api.md + a CI negative-test step assert the validator rejects obsolete APIs. The check caught real rot fixed here: two cookbook snippets used the obsolete TextView (now a plain content View, pointing at gui-cs/Editor's EditorView). Result: 29 compiled, 1 skipped, 0 failed; negative test green. Rebased onto develop after #5478 merged (squash), keeping only this PR's files; no overlap with the merged agent-docs lint changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Fix Validate Doc Snippets CI: full history for GitVersion The workflow built Terminal.Gui without fetch-depth:0, so GitVersion.MsBuild failed on the shallow clone ("Repository is a shallow clone"). Match the build-validation workflow's checkout (fetch-depth: 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Notify gui-cs/Editor on NuGet publish (#5483) Dispatches `terminal-gui-published` (payload: the published SemVer) to gui-cs/Editor after each publish. Editor's bump-terminal-gui.yml picks it up, pins the new version, validates with its full test suites, and republishes Editor against it - making Editor a continuous canary for TG API churn (see Editor specs/decisions.md DEC-010). No-op until an EDITOR_DISPATCH_TOKEN secret (PAT with repo scope on gui-cs/Editor) is configured; Editor also polls NuGet on a 6h schedule as a fallback, so nothing breaks without it. Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * Fix publish.yml: secrets context is invalid in step if conditions The Editor-notify guard from #5483 used `if: secrets.X != ''`, which is not an available context for step ifs and invalidated the whole workflow (publish run 27422643812). Mirror the secret into job-level env and gate on that, same pattern as the TEMPLATE/clet dispatches in gui-cs/Editor. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * Mandelbrot: fill display with image view + regenerate GIF with Kitty graphics (#5511) * Regenerate Mandelbrot.gif with Kitty graphics rendering The previous capture forced the Sixel raster protocol because agg could not render Kitty graphics. tuirec v0.9.0 now defaults to Kitty graphics and its pinned agg (v1.11.0-sixel, built on a Kitty-capable avt) renders them, so the Mandelbrot scenario records through its preferred Kitty path. Recorded with the released tuirec + auto-downloaded agg. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Mandelbrot: fill the display with the image view; richer zoom/pan demo Anchor the MandelbrotImageView at (0,0) and size it Dim.Fill() so the fractal fills the whole display region instead of a small centered box, and move the status label to the bottom edge. Re-record the docfx GIF with a smoother tour: a multi-step zoom into the seahorse valley, a pan across the seahorse filaments, and a zoom back out — demonstrating the smooth zooming and panning the larger image view makes possible. * Scripts/tuirec: document Kitty graphics as the default raster path tuirec v0.9.0 advertises a Kitty identity by default, so raster apps (Mandelbrot, Images) now record via Kitty graphics, which agg v1.11.0-sixel renders. Add a 'Raster graphics' section covering protocol selection, how to confirm which protocol the cast captured, that the #84 cell-size mismatch is sixel-only, and a smooth zoom/pan recording tip. Update the validation checklist and troubleshooting for Kitty + sixel. * Scripts/tuirec: add exact one-shot recipe for Mandelbrot.gif Capture the precise keystroke tour, flags, and the design rationale (seahorse-valley target coords, center-anchored keyboard zoom, the span ~1.0 / 80-iteration black-out limit, fill layout) so a future agent can reproduce the hero GIF in one shot instead of re-deriving it. * Scripts/tuirec: cross-platform commands + runnable verification Address feedback from an agent that ran the guide on Linux: - Add bash equivalents alongside PowerShell for the raster recipes, validation greps, and error check (the raster recipes are Linux/macOS only, but every snippet was PowerShell). Make the Mandelbrot recipe bash-primary. - Add a fresh-container bootstrap (dotnet SDK matching global.json, Go install, PATH for ~/.dotnet and GOPATH/bin) and fix the Windows-only GOPATH PATH note. - Give a python3/Pillow one-liner to extract a mid-zoom frame, plus the expected readout (Center X ~ -0.74, Y ~ 0.105, Span ~ 1.0) so the spot-check measures a landmark instead of a vibe. - Soften exact payload counts to robust invariants (thousands of Kitty, zero sixel, ~0.9 MB) since the count drifts with timing/font-size. - Note the '#84 font-size adjust' log line is expected and auto-corrected in v0.9.0, harmless for the Kitty path. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: Kevin Harder <kevin@kevinharder.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
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
Adds
Scripts/DocSnippetValidator— a Roslyn-based tool that compiles every fenced C# block in the agent-facing docs (ai-v2-primer.md,.claude/tasks/build-app.md,.claude/cookbook/common-patterns.md) against the freshly builtTerminal.Gui.dll— plus a CI workflow that runs it when either the docs or the library change, so API renames break loudly instead of silently rotting the examples agents copy from.What the first run found
14 of 30 snippets did not compile.
ai-v2-primer.mdwas fully green; the cookbook had substantial v1 rot. All fixed against the current API:ColorScheme = Colors.ColorSchemes ["Error"]SchemeName = "Error"new MenuItem ("New", "", action, null, null, KeyCode.N | KeyCode.CtrlMask)new MenuItem { Title = "_New", Key = Key.N.WithCtrl, Action = ... }TileViewViewArrangement.RightResizableTabView/TabTabs—Addviews,Titleis the tab label,ValueselectslistView.SelectedItemChangedValueChanged(IValue<int?>);SelectedItemisint?tableView.SelectedCellChanged/SelectedRowValueChanged(IValue<TableSelection?>),SelectedCellPointoverride OnLoaded ()onRunnableOnIsRunningChanged (bool)OpenDialog.DirectoryPath, settableSaveDialog.FileNamePath;FileNameis read-onlyListWrapper<string> (List<string>)ObservableCollection<T>int result = MessageBox.Query (...)int?view.AddCommand (...)on an instanceAddCommandis protected — call inside the subclassAlso of note: the compressed type index in AGENTS.md claims
TableViewhasSelectedRow/SelectedColumn— it does not. Worth regenerating that index in a follow-up.How it works
Runnable<string?>harness with common fields (app,view,button, ...) in scope; fragments that declare methods are retried as class members.// WRONG/❌/✗are skipped automatically;<!-- snippet: ignore -->opts a block out explicitly.Testing
dotnet run --project Scripts/DocSnippetValidator -- Terminal.Gui/bin/Debug/net10.0/Terminal.Gui.dll ai-v2-primer.md .claude/tasks/build-app.md .claude/cookbook/common-patterns.md→ 29 compiled, 1 skipped, 0 failed (and correctly reported all 14 failures before the doc fixes).🤖 Generated with Claude Code