Commit Graph

11 Commits

Author SHA1 Message Date
jaeone94
a1e6fb36d2 refactor: harden ChangeTracker lifecycle with self-defending API (#10816)
## Summary

Harden the `ChangeTracker` lifecycle to eliminate the class of bugs
where an inactive workflow's tracker silently captures the wrong graph
state. Renames `checkState()` to `captureCanvasState()` with a
self-defending assertion, introduces `deactivate()` and
`prepareForSave()` lifecycle methods, and closes a latent undo-history
corruption bug discovered during code review.

## Background

ComfyUI supports multiple workflows open as tabs, but only one canvas
(`app.rootGraph`) exists at a time. When the user switches tabs, the old
workflow's graph is unloaded and the new one is loaded into this shared
canvas.

The old `checkState()` method serialized `app.rootGraph` into
`activeState` to track changes for undo/redo. It had no awareness of
*which* workflow it belonged to -- if called on an inactive tab's
tracker, it would capture the active tab's graph data and silently
overwrite the inactive workflow's state. This caused permanent data loss
(fixed in PR #10745 with caller-side `isActive` guards).

The caller-side guards were fragile: every new call site had to remember
to add the guard, and forgetting would reintroduce the same silent data
corruption. Additionally, `beforeLoadNewGraph` only called `store()`
(viewport/outputs) without `checkState()`, meaning canvas state could be
stale if a tab switch happened without a preceding mouseup event.

### Before (fragile)

```
saveWorkflow(workflow):
  if (isActive(workflow))              <-- caller must remember this guard
    workflow.changeTracker.checkState()      <-- name implies "read", actually writes
  ...

beforeLoadNewGraph():
  activeWorkflow.changeTracker.store()      <-- only saves viewport, NOT graph state
```

### After (self-defending)

```
saveWorkflow(workflow):
  workflow.changeTracker.prepareForSave()   <-- handles active/inactive internally
  ...

beforeLoadNewGraph():
  activeWorkflow.changeTracker.deactivate() <-- captures graph + viewport together
```

## Changes

- Rename `checkState` to `captureCanvasState` with active-tracker
assertion
- Add `deactivate()` and `prepareForSave()` lifecycle methods
- Fix undo-history corruption: `captureCanvasState()` guarded by
`_restoringState`
- Fix viewport regression during undo: `deactivate()` skips
`captureCanvasState()` during undo/redo but always calls `store()` to
preserve viewport (regression from PR #10247)
- Log inactive tracker warnings unconditionally at warn level (not
DEV-only)
- Deprecated `checkState()` wrapper for extension compatibility
- Rename `checkState` to `captureCanvasState` in
`useWidgetSelectActions` composable
- Add `appModeStore.ts` to manual call sites documentation
- Add `checkState()` deprecation note to architecture docs
- Add 16 unit tests covering all guard conditions, lifecycle methods,
and undo behavior
- Add E2E test: "Undo preserves viewport offset"

## New ChangeTracker Public API

| Method | Caller | Purpose |
|--------|--------|---------|
| `captureCanvasState()` | Event handlers, UI interactions | Snapshots
canvas into activeState, pushes undo. Asserts active tracker. |
| `deactivate()` | `beforeLoadNewGraph` only | `captureCanvasState()`
(skipped during undo/redo) + `store()`. Freezes state for tab switch. |
| `prepareForSave()` | Save paths only | Active: `captureCanvasState()`.
Inactive: no-op. |
| `checkState()` | **Deprecated** -- extensions only | Wrapper that
delegates to `captureCanvasState()` with deprecation warning. |
| `store()` | Internal to `deactivate()` | Saves viewport, outputs,
subgraph navigation. |
| `restore()` | `afterLoadNewGraph` | Restores viewport, outputs,
subgraph navigation. |
| `reset()` | `afterLoadNewGraph`, save | Resets initial state (marks as
"clean"). |

## Test plan

- [x] Unit tests: 16 tests covering all guard conditions, state capture,
undo queue behavior
- [x] E2E test: "Undo preserves viewport offset" verifies no viewport
drift on undo
- [x] E2E test: "Prevents captureCanvasState from corrupting workflow
state during tab switch"
- [x] Existing E2E: "Closing an inactive tab with save preserves its own
content"
- [ ] Manual: rapidly switch tabs during undo/redo, verify no viewport
drift
- [ ] Manual: verify extensions calling `checkState()` see deprecation
warning in console
2026-04-16 12:54:12 +00:00
pythongosssss
ecb6fbe8fb test: Add waitForWorkflowIdle & remove redundant nextFrame (#11264)
## Summary

More cleanup and reliability

## Changes

- **What**: 
- Add wait for idle
- Remove redundant nextFrames

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-11264-test-Add-waitForWorkflowIdle-remove-redundant-nextFrame-3436d73d3650812c837ac7503ce0947b)
by [Unito](https://www.unito.io)

Co-authored-by: Alexander Brown <drjkl@comfy.org>
2026-04-15 16:52:41 +00:00
Alexander Brown
52ccd9ed1a refactor: internalize nextFrame() into fixture/helper methods (#11166)
## Summary

Internalize `nextFrame()` calls into fixture/helper methods so spec
authors don't need to remember to call it after common operations.
`nextFrame()` waits for one `requestAnimationFrame` (~16ms) — an extra
call is always safe, making this a low-risk refactor.

## Changes

### Phase 1: `SettingsHelper.setSetting()`
`setSetting()` now calls `nextFrame()` internally. Removed 15 redundant
calls across 7 files.

### Phase 2: `CommandHelper.executeCommand()`
`executeCommand()` now calls `nextFrame()` internally. Removed 15
redundant calls across 7 files, including the now-redundant call in
`AppModeHelper.toggleAppMode()`.

### Phase 3: `WorkflowHelper.loadGraphData()`
New helper wraps `page.evaluate(loadGraphData)` + `nextFrame()`.
Migrated `SubgraphHelper.serializeAndReload()` and `groupNode.spec.ts`.

### Phase 4: `NodeReference` cleanup
Removed redundant `nextFrame()` from `copy()`, `convertToGroupNode()`,
`resizeNode()`, `dragTextEncodeNode2()`, and
`convertDefaultKSamplerToSubgraph()`. Removed 6 spec-level calls after
`node.click('title')`.

### Phase 5: `KeyboardHelper.press()` and `delete()`
New convenience methods that press a key and wait one frame. Converted
40 `canvas.press(key)` + `nextFrame()` pairs across 13 spec files.

### Phase 6: `ComfyPage.expectScreenshot()`
New helper combines `nextFrame()` + `toHaveScreenshot()`. Converted 45
pairs across 12 spec files.

## Total impact
- **~130 redundant `nextFrame()` calls eliminated** across ~35
spec/helper files
- **3 new helper methods** added (`loadGraphData`, `press`/`delete`,
`expectScreenshot`)
- **2 existing methods** enhanced (`setSetting`, `executeCommand`)

## What was NOT changed
- `performance.spec.ts` frame-counting loops (intentional)
- `ComfyMouse.ts` / `CanvasHelper.ts` (already internalized)
- `SubgraphHelper.packAllInteriorNodes()` (deliberate orchestration)
- Builder helpers (already internalized)

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-11166-refactor-internalize-nextFrame-into-fixture-helper-methods-33f6d73d3650817bb5f6fb46e396085e)
by [Unito](https://www.unito.io)

---------

Co-authored-by: Amp <amp@ampcode.com>
2026-04-15 15:25:47 +00:00
Alexander Brown
f1bb756929 fix: remove redundant and counterproductive e2e timeout overrides (#11110)
## Summary

Alright, alright, alright. These e2e tests have been runnin' around like
they're late for somethin', settin' tight little timeouts like the
world's gonna end in 250 milliseconds. Man, you gotta *breathe*. Let the
framework do its thing. Go slow to go fast, that's what I always say.

## Changes

- **What**: Removed ~120 redundant timeout overrides from auto-retrying
Playwright assertions (`toBeVisible`, `toBeHidden`, `toHaveCount`,
`toBeEnabled`, `toHaveAttribute`, `toContainText`, `expect.poll`) where
5000ms is already the default. Also removed sub-5s timeouts (1s, 2s, 3s)
that were just *begging* for flaky failures — like wearin' a belt and
suspenders and also holdin' your pants up with both hands. Raised the
absurdly short timeouts in `customMatchers.ts` (250ms `toPass` → 5000ms,
256ms poll → default). Kept `timeout: 5000` on `.toPass()` calls
(defaults to 0), `.waitFor()`, `waitForRequest`, `waitForFunction`,
intentionally-short timeouts inside retry loops, and conditional
`.isVisible()/.catch()` checks — those fellas actually need the help.

## Review Focus

Every remaining timeout in the diff is there for a *reason*. The ones on
`.toPass()` stay because that API defaults to zero — it won't retry at
all without one. The ones on `.waitFor()` and `waitForRequest` stay
because those are locator actions, not auto-retrying assertions. The
intentionally-short ones inside `toPass` retry loops
(`interaction.spec.ts`) and the negative assertions (`actionbar.spec.ts`
confirming no response arrives) — those are *supposed* to be tight.

The short timeouts on regular assertions were actively *encouragin'*
flaky failures. That's like settin' your alarm for 4 AM and then gettin'
mad you're tired. Just... don't do that, man. Let things take the time
they need.

38 files, net -115 lines. Less code, more chill. That's livin'.

---------

Co-authored-by: Amp <amp@ampcode.com>
2026-04-10 19:47:20 +00:00
Alexander Brown
0132c77c7d test: harden 82 Playwright specs for deterministic CI runs (#10967)
## Summary

Harden 98 E2E spec files and 8 fixtures/helpers for deterministic CI
runs by replacing race-prone patterns with retry-safe alternatives.

No source code changes -- only `browser_tests/` is touched.

## Changes

- **E2E spec hardening** (98 spec files, 6 fixtures, 2 helpers):

  | Fix class | Sites | Examples |
  |-----------|-------|---------:|
| `expect(await ...)` -> `expect.poll()` | ~153 | interaction,
defaultKeybindings, workflows, featureFlags |
| `const x = await loc.count(); expect(x)` -> `toHaveCount()` | ~19 |
menu, linkInteraction, assets, bottomPanelShortcuts |
| `nextFrame()` -> `waitForHidden()` after menu clicks | ~22 |
contextMenu, rightClickMenu, subgraphHelper |
| Redundant `nextFrame()` removed | many | defaultKeybindings, minimap,
builderSaveFlow |
| `expect(async () => { ... }).toPass()` retry blocks | 5 | interaction
(graphdialog dismiss guard) |
| `force:true` removed from `BaseDialog.close()` | 1 | BaseDialog
fixture |
| ContextMenu `waitForHidden` simplified (check-then-act race removed) |
1 | ContextMenu fixture |
| Non-deterministic node order -> proximity-based selection | 1 |
interaction (toggle dom widget) |
  | Tight poll timeout (250ms) -> >=2000ms | 2 | templates |

- **Helper improvements**: Exposed locator getters on
`ComfyPage.domWidgets`, `ToastHelper.toastErrors`, and
`WorkflowsSidebarTab.activeWorkflowLabel` so callers can use retrying
assertions (`toHaveCount()`, `toHaveText()`) directly.

- **Flake pattern catalog**: Added section 7 table to
`browser_tests/FLAKE_PREVENTION_RULES.md` documenting 8 pattern classes
for reviewers and future authors.

- **Docs**: Fixed bad examples in `browser_tests/README.md` to use
`expect.poll()`.

- **Breaking**: None
- **Dependencies**: None

## Review Focus

- All fixes follow the rules in
`browser_tests/FLAKE_PREVENTION_RULES.md`
- No behavioral changes to tests -- only timing/retry strategy is
updated
- The `ContextMenu.waitForHidden` simplification removes a
swallowed-error anti-pattern; both locators now use direct `waitFor({
state: 'hidden' })`

---------

Co-authored-by: Amp <amp@ampcode.com>
Co-authored-by: github-actions <github-actions@github.com>
2026-04-09 20:50:56 -07:00
Alexander Brown
3b78dfbe1c test: migrate browser_tests/ to @e2e/ path alias and add lint rule (#10958)
## Summary

Complete the @e2e/ path alias migration started in #10735 by converting
all 354 remaining relative imports and adding a lint rule to prevent
backsliding.

## Changes

- **What**: Migrate all relative imports in browser_tests/ to use
`@e2e/` (intra-directory) and `@/` (src/ imports) path aliases. Add
`no-restricted-imports` ESLint rule banning `./` and `../` imports in
`browser_tests/**/*.ts`. Suppress pre-existing oxlint `no-eval` and
`no-console` warnings exposed by touching those files.

## Review Focus

- ESLint flat-config merging: the `@playwright/test` ban and
relative-import ban are in two separate blocks to avoid last-match-wins
collision with the `useI18n`/`useVirtualList` blocks higher in the
config.
- The `['./**', '../**']` glob patterns (not `['./*', '../*']`) are
needed to catch multi-level relative paths like `../../../src/foo`.

Follows up on #10735

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10958-test-migrate-browser_tests-to-e2e-path-alias-and-add-lint-rule-33c6d73d365081649d1be771eac986fd)
by [Unito](https://www.unito.io)

Co-authored-by: Amp <amp@ampcode.com>
2026-04-08 11:28:59 -07:00
Alexander Brown
4cb83353cb test: stabilize flaky Playwright tests (#10817)
Stabilize flaky Playwright tests by improving test reliability.

This PR aims to identify and fix flaky e2e tests.

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10817-test-stabilize-flaky-Playwright-tests-3366d73d365081ada40de73ce11af625)
by [Unito](https://www.unito.io)

---------

Co-authored-by: Amp <amp@ampcode.com>
2026-04-07 19:47:27 -07:00
Dante
ccdde8697c test: add E2E regression tests for workflow tab save bug (#10815)
## Summary

- Adds E2E regression tests for the bug fixed in PR #10745 where closing
an inactive workflow tab would save the active workflow's content into
the closing tab's file
- Three test scenarios covering the full range of the bug surface:
  1. Closing an unmodified inactive tab preserves both workflows
2. Closing a modified inactive tab with "Save" preserves its own content
(not the active tab's)
3. Closing an unsaved inactive tab with "Save As" preserves its own
content

## Linked Issues

- Regression coverage for #10745 / Comfy-Org/ComfyUI#13230

## Test plan

- [ ] `pnpm exec playwright test
browser_tests/tests/topbar/workflowTabSave.spec.ts` passes against the
current main (with the fix)

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10815-test-add-E2E-regression-tests-for-workflow-tab-save-bug-3366d73d365081eab409ed303620a959)
by [Unito](https://www.unito.io)

---------

Co-authored-by: GitHub Action <action@github.com>
2026-04-08 10:40:12 +09:00
Dante
0d535631a5 refactor(test): extract dialog page objects from inline getByRole usage (#10822)
## Summary

Extract inline `getByRole('dialog')` calls across E2E tests into
reusable page objects.

## Changes

- **What**: Extract `ConfirmDialog` class from `ComfyPage.ts` into
`browser_tests/fixtures/components/ConfirmDialog.ts` with new `save`
button locator. Add `MediaLightbox` and `TemplatesDialog` page objects.
Refactor 4 test files to use these page objects instead of raw dialog
locators.
- **Skipped**: `appModeDropdownClipping.spec.ts` uses
`getByRole('dialog')` for a PrimeVue Popover (not a true dialog), left
as-is.

## Review Focus

The `ConfirmDialog.click()` method now supports a `save` action used by
`workflowPersistence.spec.ts`, which also waits for the dialog mask to
disappear and workflow service to settle.

Fixes #10723

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10822-refactor-test-extract-dialog-page-objects-from-inline-getByRole-usage-3366d73d365081b3bc0ee7ef0ddce658)
by [Unito](https://www.unito.io)
2026-04-06 09:49:32 -07:00
jaeone94
e11a1776ed fix: prevent saving active workflow content to inactive tab on close (#10745)
## Summary

- Closing an inactive workflow tab and clicking "Save" overwrites that
workflow with the **active** tab's content, causing permanent data loss
- `saveWorkflow()` and `saveWorkflowAs()` call `checkState()` which
serializes `app.rootGraph` (the active canvas) into the inactive
workflow's `changeTracker.activeState`
- Guard `checkState()` to only run when the workflow being saved is the
active one — in both `saveWorkflow` and `saveWorkflowAs`

## Linked Issues

- Fixes https://github.com/Comfy-Org/ComfyUI/issues/13230

## Root Cause

PR #9137 (commit `9fb93a5b0`, v1.41.7) added
`workflow.changeTracker?.checkState()` inside `saveWorkflow()` and
`saveWorkflowAs()`. `checkState()` always serializes `app.rootGraph` —
the graph on the canvas. When called on an inactive tab's change
tracker, it captures the active tab's data instead.

## Test plan

- [x] E2E: "Closing an inactive tab with save preserves its own content"
— persisted workflow B with added node, close while A is active, re-open
and verify
- [x] E2E: "Closing an inactive unsaved tab with save preserves its own
content" — temporary workflow B with added node, close while A is
active, save-as with filename, re-open and verify
- [x] Manual: open A and B, edit B, switch to A, close B tab, click
Save, re-open B — content should be B's not A's
2026-03-30 12:12:38 -07:00
Christian Byrne
b12b20b5ab test: add 12 workflow persistence playwright tests (#10547)
## What

12 regression tests covering 10 workflow persistence bug gaps, including
the **critical data corruption fix in PR #9531** (pythongosssss) which
previously had ZERO test coverage.

## Why

Deep scan of 37 workflow persistence bugs found 12 E2E-testable gaps
with no regression tests. Workflow persistence is a core reliability
concern — data corruption bugs are the highest risk category.

## Tests

### 🔴 Critical
| Bug | PR | Tests | Description |
|-----|----|-------|-------------|
| Data corruption | #9531 | 2 | checkState during graph loading corrupts
workflow data |
| State desync | #9533 | 2 | Rapid tab switching desyncs workflow/graph
state |

### 🟡 Medium  
| Bug | PR/Commit | Tests | Description |
|-----|-----------|-------|-------------|
| Lost previews | #9380 | 1 | Node output previews lost on tab switch |
| Stale canvas | 44bb6f13 | 1 | Canvas not cleared before loading new
workflow |
| Widget loss | #7648 | 1 | Widget values lost on graph change |
| API format | #9694 | 1 | API format workflows fail with missing nodes
|
| Paste duplication | #8259 | 1 | Middle-click paste duplicates workflow
|
| Blob URLs | #8715 | 1 | Transient blob: URLs in serialization |

### 🟢 Low
| Bug | PR/Commit | Tests | Description |
|-----|-----------|-------|-------------|
| Locale break | #8963 | 1 | Locale change breaks workflows |
| Panel drift | — | 1 | Splitter panel size drift |

## Conventions
- All tests use Vue nodes + new menu enabled
- Each test documents which PR/commit it regresses
- Proper waits (no sleeps)
- Screenshots scoped to relevant elements
- Tests read like user stories

## 🎉 Shoutout
PR #9531 by @pythongosssss was a critical data corruption fix that now
has regression test coverage for the first time.

Part of: Test Coverage Q2 Overhaul (REG-01)

┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10547-test-12-workflow-persistence-regression-tests-incl-critical-PR-9531-32f6d73d3650818796c6c5950c77f6d1)
by [Unito](https://www.unito.io)

---------

Co-authored-by: GitHub Action <action@github.com>
2026-03-29 15:57:13 -07:00