aglinxinyuan opened a new pull request, #8133:
URL: https://github.com/apache/texera/pull/8133
### What changes were proposed in this PR?
Five existing dashboard/workspace specs extended, 202 tests to 236.
| File | Codecov lines | Branch arms |
|---|---|---|
| `share-access.component.ts` | 159/169 → **169/169** | 88/98 → **98/98** |
| `share-access.component.html` | 66/73 → **73/73** | 10/10 |
| `user-model.component.ts` | 79/82 → **82/82** | 38/42 → 40/42 |
| `user-model.component.html` | 30/34 → **34/34** | 6/6 |
| `user-workflow-list-item.component.ts` | 36/41 → **41/41** | 22/24 →
**24/24** |
| `user-workflow-list-item.component.html` | 68/71 → **71/71** | 4/4 |
| `card-item.component.ts` | 181/190 → 188/190 | 95/102 → 100/102 |
| `card-item.component.html` | 103/106 → 105/106 | 10/14 → 12/14 |
| `workspace.component.ts` | 129/133 → 130/133 | 35/42 → 36/42 |
**+42 fully-covered lines and +22 branch arms** (851/913 → 893/913). Six
files reach 100%. The plain line-hit metric moves +21 — half the Codecov gain —
because many of these lines were already executed and flip only by completing a
branch arm; the two numbers are not interchangeable and both are given.
### The headline target contributes nothing, and that is the main finding
`workspace.component.html` shows **0 of 14 lines** on Codecov and was the
reason this bundle was picked. It stays at 0/14.
`workspace.component.spec.ts:164` calls
`TestBed.overrideComponent(WorkspaceComponent, { set: { imports: [], providers:
[], schemas: [CUSTOM_ELEMENTS_SCHEMA] } })`. That forces a JIT recompile, so
the istanbul-instrumented AOT template function is never executed — measured,
not inferred: the spec runs 24 tests green and every one of the 15 statements
in the template reads `hits=0`, along with the `@ViewChild` view-query and the
`beforeunload` host-binding handler at `FNDA:0`.
Unblocking it means providing each child's transitive dependencies. The
eight child components declare **62 constructor dependencies** between them
(workflow-editor 19, menu 22, result-panel 9), before their `providedIn:
'root'` services — JointJS paper init, Monaco, the websocket services.
Overriding each *child* instead would keep the parent template instrumented but
still construct each child class and run its `ngAfterViewInit` against DOM an
emptied template does not have. `vitest.config.ts` already documents this suite
OOM-ing workers at 2 GB (#7975). Refused as an accepted limitation rather than
attempted.
### A trap that would have wasted the work
Angular's generated listener source-spans **chain from the end of the
previous listener attribute**, so a zero-hit line in an `.html` file is
frequently not the handler written on that line.
`card-item.component.html:105` reads as
`(keydown.enter)="confirmUpdateCustomName(entry.name)"`, and an existing
passing test already dispatches exactly that key event — yet the line showed
zero. The `statementMap` explains it: the `keydown.enter` function is declared
at `104:56` and is already covered; line 105's zero statement is the
`(click)="$event.stopPropagation()"` handler. Every `.html` line in this PR was
planned from `--coverage-reporters=json` and the `statementMap`/`fnMap`, never
from the template source.
### Verification
**50 semantic mutants, all 50 killed.** Each applied one at a time, with a
sha256 snapshot check, an empty-production-diff check and a "no non-spec file
modified" check before every compile, reverted from the scratch snapshot and
re-verified after. The builder's 31-mutant table was re-derived from scratch
rather than carried over, with every kill credit re-read from the run log.
Two reviewers returned ten findings; all repaired. **The repair pass moved
coverage by exactly zero** — its three new tests and ~15 new assertions
exercise lines the bundle already reached. What changed is that those lines are
now *pinned* rather than merely executed, and that is stated rather than sold
as extra coverage.
One further trap confirmed and worth carrying: the `DA` line set for a file
depends on which specs are in the run, so the before and after measurements
must use an identical `--include` filter. Both runs here used the same five
spec files.
`card-item.component.ts:310` sits inside a `setTimeout` arrow body —
istanbul instruments arrow bodies, so unlike the Scala modules' `$anonfun` rule
it does count, but the test must not call `detectChanges` after
`onEditDescription()` or the `@ViewChild` resolves and the false arm is lost.
It uses `fakeAsync` + `tick(0)`.
`yarn format:ci` passes. `frontend/junit.xml` is regenerated by every run,
is **not** gitignored, and is not committed. No production file is touched.
### Any related issues, documentation, discussions?
Closes #8131
### How was this PR tested?
```
npx ng test --watch=false --include="**/share-access.component.spec.ts"
--include="**/card-item.component.spec.ts"
--include="**/user-model.component.spec.ts"
--include="**/user-workflow-list-item.component.spec.ts"
--include="**/workspace.component.spec.ts"
```
```
Test Files 5 passed (5)
```
Re-run after rebasing onto current `main`.
### Was this PR authored or co-authored using generative AI tooling?
Generated-by: Claude Code (Opus 5)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
