Files
civitai__civitai/tests/preview-generation.spec.ts
T

203 lines
9.8 KiB
TypeScript
Raw Normal View History

test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
import { expect, test } from '@playwright/test';
import { storageStatePath } from './preview-fixtures';
import { retryFlaky } from './preview-retry';
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
/**
* Generation cost-quote e2e for a deployed PR preview.
*
* De-mocks the orchestrator: instead of stubbing the price endpoint, this drives
* the REAL `/generate` page as a gate-passing PAID member (gold) and asserts the
* client actually fires the tRPC QUERY `orchestrator.whatIfFromGraph` and gets a
* numeric Buzz cost back. whatIf is a pure price quote (no Buzz balance needed),
* so it fires on-load once a model+workflow is selected — the default form
* preselects a model, so we PREFER the on-load path (no interaction).
*
* Runs only under playwright.preview.config.ts (needs PREVIEW_URL + minted
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
* storage states). Cost source of truth: the whatIf payload
* `{ allowMatureContent, transactions, cost: { …, total }, ready }` returned by
* `orchestration-new.service.ts:whatIfFromGraph` (see also useWhatIfFromGraph.ts).
* 🔴 That payload reaches the browser in THREE different wire shapes depending on
* whether the `trpcBatching` flag is on for the session — an unbatched envelope, a
* batched JSON array, or newline-delimited `application/jsonl` stream chunks. Both
* the URL matcher and the body parser below must handle all three; assuming the
* unbatched shape is what made this test fail 100% of the time once batching
* ramped (see the comments on `isWhatIfResponse` / `extractCostTotal`).
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
*/
// Use the PAID member — passes the preview gate AND is a real generation user.
test.describe('generation cost quote (gold)', () => {
test.use({ storageState: storageStatePath('gold') });
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
const WHATIF_PROCEDURE = 'orchestrator.whatIfFromGraph';
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
/**
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
* Does this response carry the whatIf query?
*
* 🔴 Do NOT go back to `url().includes('/api/trpc/orchestrator.whatIfFromGraph')`.
* With the `trpcBatching` flag on (`httpBatchStreamLink`, src/utils/trpc.ts), the
* client coalesces concurrent queries into ONE request whose path is the
* COMMA-JOINED procedure list, and whatIf is not first:
* /api/trpc/content.get,challenge.getInfinite,generationPreset.getOwn,
* wildcardSet.getMyUserSet,user.userRewardDetails,
* orchestrator.whatIfFromGraph?batch=1&input=…
* That URL does not contain the `/api/trpc/<proc>` prefix, so a prefix match never
* fires and the wait times out on a request that DID happen and DID return 200.
* Batch membership varies run to run, so match the procedure inside the path list.
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
*/
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
function isWhatIfResponse(url: string): boolean {
let pathname: string;
try {
pathname = new URL(url).pathname;
} catch {
return false;
}
const prefix = '/api/trpc/';
if (!pathname.startsWith(prefix)) return false;
return pathname.slice(prefix.length).split(',').includes(WHATIF_PROCEDURE);
}
/**
* Pull the whatIf `cost.total` out of a raw tRPC response body.
*
* The body arrives in one of three wire shapes depending on link + batching, so
* parse the RAW TEXT rather than `response.json()` (which throws on the streamed
* one):
* 1. unbatched `{"result":{"data":{"json":{…,"cost":{…},"ready":true}}}}`
* 2. batched `[{…},…,{"result":{"data":{"json":{…}}}}]`
* 3. batched + streamed — `httpBatchStreamLink` sends `trpc-accept: application/jsonl`,
* so the server answers with NEWLINE-DELIMITED JSON chunks that reference each
* other by index; `JSON.parse` of the whole body fails, and the payload sits at a
* different depth (`{"json":[17,0,[[{…,"cost":{…},"ready":true}]]]}`).
*
* The one thing stable across all three is the whatIf payload object itself, so find
* it structurally: an object with BOTH a numeric `cost.total` and a boolean `ready`.
* Requiring `ready` scopes the search to whatIf — the payload contract in
* `orchestration-new.service.ts:whatIfFromGraph` always returns both — so a sibling
* procedure sharing the batch can't satisfy the match and produce a false pass.
*/
function extractCostTotal(raw: string): number | null {
const docs: unknown[] = [];
try {
docs.push(JSON.parse(raw));
} catch {
for (const line of raw.split('\n')) {
const trimmed = line.trim();
if (!trimmed) continue;
try {
docs.push(JSON.parse(trimmed));
} catch {
// A partial/among-chunks line: skip it, another chunk carries the payload.
}
}
}
const seen = new Set<unknown>();
const walk = (value: unknown): number | null => {
if (value === null || typeof value !== 'object') return null;
if (seen.has(value)) return null;
seen.add(value);
const node = value as any;
if (typeof node?.cost?.total === 'number' && typeof node?.ready === 'boolean') {
return node.cost.total;
}
for (const child of Object.values(node as Record<string, unknown>)) {
const hit = walk(child);
if (hit !== null) return hit;
}
return null;
};
for (const doc of docs) {
const hit = walk(doc);
if (hit !== null) return hit;
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
}
return null;
}
// 1. Primary, network-based: whatIf fires on /generate load and quotes a cost.
test('whatIfFromGraph fires on /generate and returns a numeric cost', async ({ page }) => {
// /generate is the heaviest SSR page; on a cold/contended single-replica preview
// pod the load+hydrate can exceed the whatIf wait, and Playwright's test-level
// retries fire within seconds — too fast to outlast a load spike. Retry the whole
// navigate+wait with backoff so a transient spike is ridden out (the assertion
// must still pass; a sustained failure surfaces after the attempts). Extend the
// per-test timeout to fit ~2 attempts of the 45s wait + navigation + backoff.
test.setTimeout(200_000);
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
// 🔴 Buffer the whatIf body OURSELVES instead of reading `response.text()` later.
// Under `httpBatchStreamLink` the app reads the whole jsonl stream (cost included)
// and then abandons the still-open reader. Chromium reports that to Playwright as
// `net::ERR_ABORTED` and evicts the resource, so a later `response.text()` fails
// with "Protocol error (Network.getResponseBody): No data found for resource with
// given identifier" — measured on 4 of 8 /generate loads, i.e. a coin-flip flake
// that has nothing to do with the product. Routing the request lets Playwright do
// the fetch and hold the complete body, removing the race entirely.
let capturedBody: string | null = null;
await page.route(
(url) => isWhatIfResponse(url.toString()),
async (route) => {
const routed = await route.fetch();
const text = await routed.text();
if (routed.status() === 200 && capturedBody === null) capturedBody = text;
await route.fulfill({ response: routed, body: text });
}
);
const { body } = await retryFlaky(
'whatIf on /generate',
async () => {
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
capturedBody = null;
// Arm the response listener BEFORE navigating so an on-load fire isn't missed.
// 45s: whatIf is the heaviest client-side flow — page load + hydrate + form
// init + resource resolve + orchestrator round-trip must complete before it
// fires. The orchestrator itself is fast (~57ms), so when healthy this never
// approaches 45s — the budget is for a cold-pod slow window.
const whatIfResponse = page.waitForResponse(
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
(r) => isWhatIfResponse(r.url()) && r.status() === 200,
{ timeout: 45_000 }
);
const resp = await page.goto('/generate', { waitUntil: 'domcontentloaded' });
expect(resp?.status(), 'HTTP status for /generate').toBeLessThan(400);
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
// Gate must not bounce a gold (paid) member.
expect(page.url(), 'should not redirect to /login').not.toContain('/login');
expect(page.url(), 'should not redirect to /preview-restricted').not.toContain(
'/preview-restricted'
);
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
// NOTE: relies on the default /generate form preselecting a valid model+workflow
// so whatIf fires without interaction. If a future default ships with no model
// preselected this will time out — see the UI-fallback note below.
test(preview): fix whatIfFromGraph smoke test against tRPC request batching (#3541) * test(preview): fix whatIfFromGraph smoke test against tRPC request batching The `whatIfFromGraph fires on /generate` preview smoke test has failed 100% of the time on every PR. It is NOT a cold pod, a DB artifact, or a product defect: driving a live preview shows the request fires ~1.3s after navigation and returns HTTP 200 with `cost.total = 8`. Two test-side defects, both introduced by `trpcBatching` (#2946) ramping on: 1. URL predicate. `httpBatchStreamLink` coalesces concurrent queries into one request whose path is the COMMA-JOINED procedure list, and whatIf is last: /api/trpc/content.get,challenge.getInfinite,...,orchestrator.whatIfFromGraph That does not contain `/api/trpc/orchestrator.whatIfFromGraph`, so `waitForResponse` never matched and timed out on a request that succeeded. Now matched by procedure name within the path list (batched or not). 2. Body shape + capture. The client sends `trpc-accept: application/jsonl`, so the response is newline-delimited chunks — `response.json()` throws, and the payload sits at a different depth than the old walk expected. Parse raw text, handling all three wire shapes, and locate the payload structurally (numeric `cost.total` + boolean `ready`, the whatIf contract) so a sibling procedure in the batch cannot produce a false pass. The app also reads the full stream and then abandons the idle reader, which Chromium reports as `net::ERR_ABORTED` and evicts — making a later `response.text()` fail with "No data found for resource" on 4 of 8 loads. The body is now buffered by a route handler, removing that race. Verified against the live preview for PR #3535: red at origin/main (45s `waitForResponse` timeout, the exact CI signature), green after, 12/12 runs (11/12 without the internal retry), ~2s to the assertion. Mutation-checked: disabling the jsonl branch fails with this guard's own error ("cost.total parsed from whatIfFromGraph response" -> null). The 45s budget is unchanged and the test still guards the pre-spend cost quote. * fix(test): narrow capturedBody honestly instead of casting through null CI Typecheck failed: tests/preview-generation.spec.ts(177,24): error TS2352: Conversion of type 'null' to type 'string' may be a mistake `capturedBody` is declared `string | null`, but it is assigned from inside the `page.route` closure, which TS control-flow analysis cannot see. After the per-attempt reset to null, TS narrows it to `null` at the return, making `as string` an illegal null->string conversion. Replaced the `expect(...).not.toBeNull()` + cast with an explicit null check. That narrows the type honestly, removes the cast entirely, and fails with a message distinguishing the two halves: the response was observed, but the route handler did not capture its body. Verified: tsc error count 22 at origin/main and 22 here in the same environment (delta 0), with zero errors mentioning this spec. The 22 are pre-existing local artifacts (no prisma generate on NixOS, missing submodule). Why this was not caught before pushing: the fix was verified by RUNNING the Playwright spec 12 times, and Playwright transpiles without full type checking. `tests/` is in the tsconfig include set, so CI checks it — a green Playwright run is not a typecheck. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: audit <alexhurwitz.dev@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-02 13:58:00 -05:00
await whatIfResponse;
// Read the body the route handler buffered, NOT `response.text()` (see above).
// Raw text, not `.json()`: under batching the body is newline-delimited JSON
// (`trpc-accept: application/jsonl`) and `.json()` throws on it.
// A plain null-check rather than `expect(...).not.toBeNull()` plus a cast.
// `capturedBody` is assigned from inside the `page.route` closure, which
// TS control-flow analysis cannot see, so after the reset above it narrows
// to `null` here and `as string` is an illegal null->string conversion.
// Checking explicitly narrows it honestly, needs no cast, and fails with a
// message that says which half broke: the response was seen, the body wasn't.
if (capturedBody === null) {
throw new Error(
'whatIf response was observed but its body was not captured by the route handler'
);
}
return { body: capturedBody };
},
{ attempts: 2 }
test(preview): tranche-2 e2e — generation whatIf, report loop, moderation (#2470) * test(preview): tranche-2 e2e — generation whatIf, report loop, moderation Extends the preview smoke harness to the audit's highest-risk untested flows, self-seeding per-run uniquely-tagged fixtures (no shared-dev-DB collisions; the mutation isolation answer to "seed per-PR"). All report-only. - preview-trpc.ts: tiny superjson+batched tRPC client (CSRF Origin/Referer stamped) for self-seeding via page.request, + uniqueToken(). - preview-generation.spec.ts: assert the REAL orchestrator.whatIfFromGraph cost quote fires + returns a numeric cost (gold). De-mocks the pricing path; no Buzz. - preview-report.spec.ts: tester self-seeds a Post (post.create) and reports it (report.create {type:'post',reason:'TOSViolation',details:{violation}}). - preview-moderation.spec.ts: mod queues (/moderator/reports + /images) render; + a self-seeded report is actioned end-to-end (report.create -> setStatus). tRPC input shapes verified against the schema files. NOT YET live-validated — the CSRF origin gate on direct tRPC + the whatIf/mod selectors get confirmed by the next report-only preview run (deferred: generation-submit, blocked on the external Buzz service which can't be seeded via Postgres). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(preview): typecheck — reportId is number | undefined the pipeline tsc (not playwright --list) caught: reportId inferred number from report?.id, then assigned mine?.id (number | undefined) -> TS2322. Type it explicitly; the runtime expect() still guards undefined. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(preview): drop the two fragile secondary DOM checks The live run validated every core tranche-2 flow (real whatIf cost, tRPC self-seed report round-trip through the CSRF gate, mod render + report->action loop). The only two reds were optional best-effort DOM checks that duplicate passing core tests: - generation "cost near submit button": the submit button is in a gen panel collapsed by default on the preview viewport (resolves but hidden). The network whatIf assertion already covers the cost. - report "image page report affordance": the report control is behind an entity action menu, not a top-level button. report.create is already covered. Removed both rather than leave known flakes in the gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-10 19:15:53 -05:00
);
const total = extractCostTotal(body);
expect(total, 'cost.total parsed from whatIfFromGraph response').not.toBeNull();
expect(typeof total, 'cost.total is numeric').toBe('number');
expect(total as number, 'cost.total is a non-negative quote').toBeGreaterThanOrEqual(0);
});
// The real pricing path is fully covered by the network assertion above. A DOM
// cost-near-submit check was dropped: the submit button + cost live in a gen
// panel that's collapsed by default on the preview viewport (button resolves
// but is `hidden`), making it a fragile, redundant assertion.
});