fix(webview): isolate parallel mode and provider profile writes (vps2 F3)

[easonLiangWorldedtech]

Draft PR — vps2 unit F3 (mode/profile writes).

Tracking: easonLiangWorldedtech#41 (vps2 series ledger). Upstream issue: #1559 (this series' gap record; the original upstream bug is #978). Port source: the F1c-head-to-CS residual, cross-checked against upstream PR #979 (fix(webview): isolate parallel mode and provider profile writes) — closed/superseded by the vps2 series; the #41 ledger names #979 as the F3 port source.

Scope

10 files, 919 insertions, 65 deletions (measured vs stack base aa0f3b1):

• src/core/webview/ClineProvider.ts (112/6) — the webview delete-path (deleteProviderProfile + repointPersistedViewStates), loadViewState dead-mode-pin validation, the handleModeSwitch profile-write isolation (per-view upsert/activate instead of the shared write), the as-any cast drop in delegateParentAndOpenChild, and a 3-line Stryker exclusion comment (next-line CallExpression,ObjectLiteral directive) above the no-op listApiConfigMeta view-state write (two equivalent mutants; see the Mutation gate trust note).
• src/core/task/Task.ts (16/2) + Task.spec.ts (40/2) — the task-side per-view mode/profile write: mode switches route through provider.handleModeSwitch and the submitted message is kept on failure.
• src/core/tools/SwitchModeTool.ts (7/4) + switchModeTool.spec.ts (22/14) — the tool-side per-view write port.
• src/core/webview/webviewMessageHandler.spec.ts (5/0) — ported mock surface only: fix(webview): isolate parallel mode and provider profile writes #979's 4 mock fields (captureCustomModeCreated, captureModeSettingChanged, updateCustomMode, handleModeSwitch) plus the defaultModeSlug import. webviewMessageHandler.ts itself is 0/0 — see Port fidelity.
• src/core/webview/tests/ClineProvider.sticky-mode.spec.ts (72/31) — the sticky-mode spec rewrites (concurrent-switch assertions re-targeted from the shared mode key to the per-view viewStates key).
• src/core/webview/tests/ClineProvider.spec.ts (20/4) — exactly the two handleModeSwitch assertion-rewrite hunks (L3328, L3362 vs base).
• src/core/webview/tests/ClineProvider.parallelMode.spec.ts (623/0) — 3 import lines + a 620-line append to the file created by F2: the CS F3 describes (310 lines, byte-identical to CS L1364-1673), 116 lines of F3-authored mutation-killing tests (corrupt-entry repoint skip, legacy-pin repoint, unknown-mode-slug no-op, released-provider switch tool path, ModeChanged event emission, empty-profile no-op), and a 194-line append of 5 further mutation-killing tests (no-task switch race, non-focused profile-load skip, upsert-without-activation list refresh, and the two deleteProviderProfile pin-behavior tests).
• src/eslint-suppressions.json (2/2) — no-explicit-any counts decrease for core/webview/ClineProvider.ts (12 to 11) and core/webview/tests/ClineProvider.sticky-mode.spec.ts (36 to 33); no count increases.

Budget

• a+d 984 vs the 400-soft / 1000-hard budget: soft-over, hard-under. The unit carries the production isolation logic (CP 112/6, Task 16/2, SwitchModeTool 7/4), three spec rewrites (sticky 72/31, CP.spec 20/4, switchModeTool.spec 22/14), and the 623-line parallelMode append. The 310 authored test lines (116 + 194) are gate-mandated: the Stryker diff gate cannot be satisfied by the CS-residual tests alone (several CP conditional/object-literal/string-literal mutants and one NoCoverage have no CS-side killer), and the skill rule is to kill by spec, never suppress. The only non-test production addition is the 3-line Stryker exclusion comment (equivalent mutants, documented in the Mutation gate trust note).
• Stryker-diff gate (vs aa0f3b1, final head 8da5c6e): 63 mutants generated on the 78 changed executable lines — 61 killed, 2 ignored (the excluded equivalent pair at the listApiConfigMeta view-state write), 0 survived, 0 uncovered; gate exit 0.
• vitest: 475/475 passed (138 suites, 16 spec files) — per file: ClineProvider.spec 206, Task.spec 101, ClineProvider.parallelMode 24, Task.persistence 24, ClineProvider.sticky-mode 19, Task.throttle 19, ClineProvider.taskHistory 18, ClineProvider.apiHandlerRebuild 15, ClineProvider.sticky-profile 15, ClineProvider.flicker-free-cancel 10, Task.pending-action 9, Task.dispose 7, ClineProvider.pending-action 4, ClineProvider.lockApiConfig 2, Task.resume-eviction-race 1, Task.sticky-profile-race 1.
• check-types, eslint (--prune-suppressions, max-warnings 0; suppression counts flat-or-down), prettier (--end-of-line=auto): all pass — check-types exit 0; eslint exit 0 on the touched files (suppression file unchanged from the stack base: no-explicit-any counts down on the two listed files, flat-or-down invariant holds); prettier --end-of-line=auto clean.

Mutation gate trust note

• Gate of record: the final sandbox-mode run at head 8da5c6e (vs stack base aa0f3b1; 78 changed executable lines; the 16-file direct test set): 63 mutants generated — 61 killed, 2 ignored, 0 survived, 0 uncovered; gate exit 0.
• The 2 ignored mutants are the excluded pair at the createNewApiConfiguration else-branch call this._updateViewLocalStateFromMutation({ listApiConfigMeta }) at L2238: the CallExpression mutant (statement deleted) and the ObjectLiteral mutant (object emptied). Both are true equivalent mutants: the method (byte-identical to the CS method) applies only mode, currentApiConfigName, apiConfiguration and provider-settings keys, and a listApiConfigMeta-only payload matches none of them, so the deleted-statement and empty-object forms are behaviorally indistinguishable from the original. The exclusion is the narrowest mutator-specific directive the gate accepts (next-line, two named mutators, concrete reason committed at the call site); broad exclusions are rejected by the gate.
• Four mutants that surfaced as Survived in earlier sandbox-mode runs on this machine — L2053 ConditionalExpression, L2072 ConditionalExpression, L2275 ConditionalExpression and L2283 ObjectLiteral — are all killed in the final run. The earlier survivors were an environment artifact (sandbox-mode per-mutant verification of ClineProvider.ts running against an un-mutated file copy in this worktree setup, while the Task.ts and SwitchModeTool.ts mutants verified and killed correctly in the same runs); the artifact was reproduced and isolated in a detached probe worktree. Killability of record is the plain-vitest true-form matrix over the committed 16-file suite (475 tests): warm baseline 475/475; L2053 true form 1 failure (the no-task switch-race test); L2072 true form 1 failure (the non-focused profile-load skip test); the L2234 string-literal form 1 failure (the upsert-without-activation list-refresh test); the L2238 excluded forms 0 failures (equivalent); L2275 true form 3 failures (the committed replacement-profile sync test plus both deleteProviderProfile pin-behavior tests); L2283 form 2 failures (both pin tests); no test fails under any mutant other than its designed killer(s).
• In-place run (STRYKER_IN_PLACE=true, same selectors and test set): not run — the canonical sandbox-mode gate passed at the final head, so a separate in-place run adds no evidence.

Port fidelity (coordinator-verified against the vps2 CS e9a44b2)

• The F3 residual (CS vs F1c head, per file) was ported hunk-by-hunk; no whole-file copies.
• webviewMessageHandler.ts is 0/0 vs base. The residual's kimi-code OAuth try/catch hunk is CS-only divergence (it is absent from upstream fix(webview): isolate parallel mode and provider profile writes #979, whose WMH.ts carries only the two F1c-shipped hunks) — CS not-ported register item 1, deliberately not ported. webviewMessageHandler.spec.ts carries only fix(webview): isolate parallel mode and provider profile writes #979's mock-field additions (5/0); the stack-side legacy-repair test, the Key-aware getValue mock, and the em-dash comment are retained from the F2 base (register item 4). fix(webview): isolate parallel mode and provider profile writes #979's mode-routing WMH.spec describe ("routes mode messages through handleModeSwitch instead of writing ContextProxy directly") exists in neither the CS nor the stack and is ported by no unit of this series — logged as a parked open question, not a gap of this PR.
• ClineProvider.ts setViewStateId proto guard: the residual removes the guard (CS lineage); this PR retains the guard exactly as shipped by F1a, so the guard region is byte-identical to the F2 head (no diff hunk there). Rationale: without the guard, states["proto"] invokes the Object.prototype setter and the durable write for that view is silently dropped; no F3 spec depends on the unguarded behavior. Cross-check: upstream fix(webview): isolate parallel mode and provider profile writes #979's setViewStateId body also has no guard, confirming the guard is stack-side (F1a) hardening, not fix(webview): isolate parallel mode and provider profile writes #979/CS content.
• ClineProvider.spec.ts: exactly the two handleModeSwitch assertion-rewrite hunks ported (each re-targets the shared-key assertion to the per-view viewStates objectContaining form, with the view pinned via setViewStateId); the F2-head describe("setViewStateId") guard test block is retained byte-identical (invisible in the diff). The remaining CS residual on this file is the deliberate F1-series describe restructure — excluded, not F3 content.
• ClineProvider.sticky-mode.spec.ts: all 4 CS hunks present at the F3 head (including the concurrent-switch filter rewrite from the shared mode key to the per-view viewStates key) — verified byte-identical to the CS residual by the coordinator.
• ClineProvider.parallelMode.spec.ts: the 310-line CS F3 describe region (CS L1364-1673) is present byte-identical in order at the F3 head; the 116 authored lines are targeted mutation killers, added in 3 blocks inside the CS describes; CS L1674 (the blank separator before the F4 describes) is not ported — F4 appends its own block.
• The defaultModeSlug import (CS register item 5) lands in this unit via the webviewMessageHandler.spec.ts mock surface (F1c had dropped it as unused).

Structural note on the parallelMode spec (coordinator-verified)

The CS parallelMode.spec.ts is 1790 lines: a shared preamble (L1-672), an F1-series test section (L673-1363: viewId uniqueness, local state isolation, saveViewState, stale temporary-id load), the F3 describes (L1364-1673), a blank separator (L1674), and the F4 multi-instance describes (L1675-1790). In this series the F1-series section lives in ClineProvider.spec.ts (shipped by F1a/F1b/F1c — the deliberate F1-series describe restructure), and the F2 unit shipped the persisted-pruning and #1065 retention tests inside the parallelMode file (they are absent from the CS parallelMode file). Both placements are behaviorally covered; the divergence is structural, not a coverage gap.

Series mechanics

• Base of record: upstream/main @ 0d937c0; PR base is main; the branch is stacked on the F2 head aa0f3b1 (the F1-series and F2 heads are merged below it in the series merge order).
• Draft PR per unit; merge order F1a to F1b to F1c to F2 to F3 to F4 to F5 to F6 to F7.
• CS not-ported register (for consistency): (1) kimi-code OAuth try/catch + routerModels.spec.ts +29 — also the reason WMH.ts is 0/0 here; (2) ApiConfigManager.tsx min-w-0 shrink to grow; (3) ApiConfigManager.visual.tsx deletion + 2 PNG baselines; (4) mojibake comment; (5) unused defaultModeSlug import — resolved by this unit; (6) providers/, .coderabbit.yaml, .github/, CONTRIBUTING.md, .gitignore churn.
• Merge check against upstream main 4c7474d (merge-base = base of record 0d937c0), via git merge-tree on the full F0-to-F3 stack: nine files auto-merge, including the four files changed on both sides (Task.ts, Task.spec.ts, ClineProvider.ts, ClineProvider.spec.ts). The single content conflict is src/eslint-suppressions.json: upstream main independently decreased six no-explicit-any counts (including core/webview/ClineProvider.ts 12 to 8 and core/task/Task.ts 19 to 17) while this PR decreases two (core/webview/ClineProvider.ts 12 to 11 and the sticky-mode spec 36 to 33, vs stack base). The one overlapping entry is core/webview/ClineProvider.ts; resolution is mechanical: regenerate that entry from the post-merge file with eslint --prune-suppressions (per the AGENTS.md invariant the count can never increase; the post-merge count is flat-or-down vs upstream 8). Every other file entry touched by either side is disjoint.

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• coderabbit-review-active

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Team

Run ID: a3c8f55e-8859-4a9d-95c7-c5c904b79df2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Review status

Thanks for contributing. This comment tracks the review sequence and the next action.

Current step: Fix the failing required CI checks; awaiting-maintainer requires CI and automated review completion.

Review-state labels are managed by this workflow; do not edit them manually.

[codecov]

Codecov Report

❌ Patch coverage is 97.83951% with 7 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/core/webview/ClineProvider.ts	98.33%	0 Missing and 4 partials ⚠️
webview-ui/src/utils/vscode.ts	86.36%	1 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!