fix(SDK-6463): robust TS config compile in NX monorepos + tolerate a11y afterEach scan timeouts by Bhargavi-BS · Pull Request #1131 · browserstack/browserstack-cypress-cli

Bhargavi-BS · 2026-06-19T09:22:05Z

SDK-6463 — Cypress test-count / status inconsistencies (customer: Sapiens, NX monorepo)

Fixes two client-side browserstack-cypress-cli defects surfaced by SDK-6463. The customer's reported dashboard count discrepancy is largely not a CLI bug (see RCA below); these two fixes address the genuine CLI defects: silently broken parallelization in NX monorepos, and accessibility-hook timeouts skipping tests.

RCA — the "inconsistency" is three layers

A. Test Observability rendering (grep-filtered tests shown as "Pending", timeout placeholders, new-dashboard counting) — faithful; owned by the TO/dashboard team, not the CLI.
B. Client-side test selection (@cypress/grep tags + browserstack.json exclude globs) — expected behavior.
C. Two real CLI defects — fixed here.

C.1 — TS cypress-config compile fails silently in NX monorepos → parallelization collapses

readCypressConfigUtil.convertTsConfig compiles the TS config to read it. In an NX/monorepo:

the extends temp tsconfig inherited base options (noEmit / emitDeclarationOnly / composite / noEmitOnError) that suppress or redirect the compiled JS, and
any tsc type error made tsc exit non-zero, which short-circuited the &&-chained tsc-alias, leaving path aliases (e.g. @org/lib) un-rewritten so the compiled config couldn't be require()d.

The read error was then silently swallowed (readCypressConfigFile catch) → cypressConfigFile = {} → getNumberOfSpecFiles fell back to the default cypress/e2e/** glob → 0 specs → setParallels forced parallels to browserCombinations.length. This is the customer's Parallels limit specified: 1 (despite parallels: 3) and the repeated Cypress config file not found at: …tmpBstackCompiledJs/… errors.

Fix (bin/helpers/readCypressConfigUtil.js):

Force emit-friendly overrides in the extends temp tsconfig: noEmit:false, emitDeclarationOnly:false, composite:false, declaration:false, declarationMap:false, noEmitOnError:false, incremental:false. These only affect the throwaway temp tsconfig used to read the config — not the user's real build.
Run tsc-alias unconditionally (& on Windows / ; on Unix instead of &&) so a tsc type error doesn't leave aliases un-rewritten. convertTsConfig already tolerates tsc errors by parsing the emitted-files output.

C.2 — accessibility `afterEach` scan/save timeout skips the rest of the spec

bin/accessibility-automation/cypress/index.js calls cy.wrap(performScan(win), {timeout:30000}) and cy.wrap(saveTestResults(win, …), {timeout:30000}) in the global afterEach with no rejection handling. When the in-page scanner doesn't echo A11Y_SCAN_FINISHED / A11Y_RESULTS_SAVED, the wrap times out, the afterEach hook fails, and Cypress skips all remaining tests in the spec — the ticket's "cy.wrap() timed out … Because this error occurred during a after each hook we are skipping all of the remaining tests" and "should have failed but got skipped".

Fix: add .catch to both wrap chains so a hung scan/save is logged, not cascaded into skipped tests.

Validation

C.1 — real BrowserStack infra (same project & config, only the CLI differs):

	Unfixed CLI	Fixed CLI
`Cypress config file not found` errors	2	0
`Parallels limit specified` (config has `parallels: 3`)	1	3
Build	`7493f8339065f0570cd0867d7bec599487bb6dd9`	`bba6009df5e214058e1567f17ac268ea312675e0`

C.1 — local NX tsconfig matrix: noEmit / emitDeclarationOnly / composite / noEmitOnError+type-error all fail to read the config before the fix and read OK after it.

C.2 — local simulation: drives the real plugin afterEach with a hung scanner; the new regression test fails without the fix and passes with it across hung-scan, hung-save, and happy-path scenarios.

Unit suite: 685 passing (was 678 → +7 new tests), 0 new failures (13 failures pre-exist on a clean tree and are unrelated).

Files

bin/helpers/readCypressConfigUtil.js — C.1 fix
bin/accessibility-automation/cypress/index.js — C.2 fix
test/unit/bin/helpers/readCypressConfigUtil.js — +3 C.1 regression tests
test/unit/bin/accessibility-automation/cypress/index.js — new, +4 C.2 regression tests

Notes / follow-ups

The dashboard count discrepancy (Layer A) should be routed to the Test Observability/dashboard team — it is not addressed here.
Latent bug spotted while investigating C.2: index.js:292 references an undefined includeTags (should be includeTagArray), reachable only when accessibility include/exclude tags are set. Not fixed here; worth a separate ticket.

🤖 Generated with Claude Code

When the cypress config is TypeScript and lives in an NX/monorepo, the extends temp tsconfig inherited base options (noEmit / emitDeclarationOnly / composite / noEmitOnError) that suppress or redirect the compiled JS, and any tsc type-error short-circuited the '&&'-chained tsc-alias so path aliases were left un-rewritten. The compiled config then could not be found/required, the error was silently swallowed, and getNumberOfSpecFiles fell back to a default glob that found 0 specs -> setParallels collapsed parallels to 1. Force emit-friendly overrides in the extends temp tsconfig and run tsc-alias unconditionally (& / ; instead of &&). Adds regression tests covering the emit-override and unconditional-alias behavior. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…K-6463) A hung accessibility scan made the 30s cy.wrap() time out and fail the afterEach hook, which makes Cypress skip ALL remaining tests in the spec (they surface as 'skipped' instead of running). Add .catch handlers to both cy.wrap(..., {timeout:30000}) chains (performScan and saveTestResults) so a timeout is logged instead of cascading into skipped tests. Adds a regression test that loads the real plugin afterEach and asserts tolerance. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

amaanbs

Automated SDK PR Review

Verdict: Request Changes — 1 finding to address before merging (W1), 1 suggestion for follow-up (S1).

Summary: 0 critical · 1 warning · 1 suggestion across 2 files reviewed.

The Unix half of the fix (; instead of && between tsc and tsc-alias) is correct and ready to ship. The Windows half (&) has a concurrency concern — on Windows CMD, & starts the second command without waiting for the first to finish, which may cause tsc-alias to process a not-yet-emitted file. This is verifiable empirically; see W1 for the idiomatic Windows fix. S1 (silent a11y scan timeout) is a follow-up suggestion, not a blocker.

See inline comments below for full Problem and Suggested Fix detail.

Generated by Automated SDK PR review.

…ways-settle (SDK-6463) The earlier afterEach used .catch() on Cypress chains (cy.wrap(...).then(...).catch(...)). Cypress Chainable has no .catch (commands are not promises), so it threw 'cy.wrap(...).then(...).catch is not a function' synchronously in the hook, failing it and aborting the rest of the spec — the customer's exact symptom. Rework: performScan/saveTestResults now ALWAYS settle (internal timeout tunable via ACCESSIBILITY_SCAN_TIMEOUT, default 25s) and guard cross-origin window access, so cy.wrap's 30s timeout never fires and the hook never throws. Verified with REAL Cypress (headed): the old code reproduces 'skipping all of the remaining tests'; the reworked code runs all tests and completes the scan path (Payload to send / Saved accessibility test results logged). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…nsion (SDK-6463) glob.sync (glob@7->minimatch@3, and even glob@10->minimatch@9) throws 'expand is not a function' / 'brace_expansion_1.default is not a function' when a project force-resolves brace-expansion to an incompatible major (e.g. 5.x) via yarn resolutions / npm overrides. That crash aborted getNumberOfSpecFiles and produced builds with 0 executed tests (or crashed the run). A secondary bug in deleteBaseUrlFromError (err.replace on a non-string) then masked the real error. - Wrap glob.sync in safeGlobSync: log once and return [] on failure so spec discovery and the run proceed (specs are resolved on BrowserStack regardless of the local count). - Backstop try/catch in getNumberOfSpecFiles so it never throws. - Guard deleteBaseUrlFromError against non-string errors. Verified against a real glob@7 + brace-expansion@5 crash. Adds regression tests. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…ion-resilience fix(SDK-6463): survive glob/minimatch crash from incompatible brace-expansion (0-test builds)

… repeated scan timeouts (SDK-6463) The always-settle rework stops scan/save promises from hanging, but the a11y afterEach can still FAIL from its own Cypress commands — e.g. cy.window() on a cross-origin page (SSO redirect to login.microsoftonline.com) or cy.task('get_test_run_uuid') when the TO node plugin isn't registered. Cypress chains have no .catch, so the correct 'catch' is the per-test fail listener: cy.on('fail', () => false) registered at the top of the hook suppresses any failure raised by the hook's commands, so the user's test is never failed and the spec is never aborted. Also adds a circuit breaker: each hung scan/save costs up to ACCESSIBILITY_SCAN_TIMEOUT (default 25s) per test; after 3 consecutive timeouts (configurable via ACCESSIBILITY_SCAN_CIRCUIT_LIMIT) the plugin stops attempting accessibility work for the rest of the spec instead of stalling every remaining test and wrapped command. Verified with REAL Cypress 13.15.0 (headed): - unregistered-task scenario: without guard -> hook fails, 'skipping all of the remaining tests' (0 passing); with guard -> all tests pass - hung-scanner scenario: all tests pass, no abort - circuit breaker: deterministic server-side count shows A11Y_SCAN stops after the circuit opens (2 hits across 4 tests) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…il-guard fix(SDK-6463): catch a11y afterEach failures (cy.on fail guard) + circuit-break repeated scan timeouts

amaanbs

Remove ticket identifiers from comments.

Addresses review feedback on #1131 — comments now describe the behavior without referencing the internal ticket. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Same review convention as requested on #1131. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Bhargavi-BS · 2026-07-06T08:19:17Z

Remove ticket identifiers from comments.

done

Bhargavi-BS and others added 2 commits June 19, 2026 14:50

Bhargavi-BS requested a review from a team as a code owner June 19, 2026 09:22

Bhargavi-BS requested review from kamal-kaur04 and pranay-v29 June 19, 2026 09:22

Merge branch 'master' into sdk-6463-nx-tsconfig-and-a11y-afterEach-fixes

94a33c2

amaanbs reviewed Jun 22, 2026

View reviewed changes

Comment thread bin/helpers/readCypressConfigUtil.js

Comment thread bin/accessibility-automation/cypress/index.js Outdated

Bhargavi-BS and others added 2 commits July 1, 2026 19:32

Bhargavi-BS mentioned this pull request Jul 1, 2026

fix(SDK-6463): survive glob/minimatch crash from incompatible brace-expansion (0-test builds) #1138

Merged

Bhargavi-BS and others added 2 commits July 2, 2026 06:15

Merge pull request #1138 from browserstack/sdk-6463-glob-brace-expans…

9b8d34b

…ion-resilience fix(SDK-6463): survive glob/minimatch crash from incompatible brace-expansion (0-test builds)

Bhargavi-BS mentioned this pull request Jul 5, 2026

perf(SDK-6463): prune excluded dirs from zip/md5 walks — minutes-long 'Creating tests.zip' stall on monorepos #1142

Open

Bhargavi-BS added 2 commits July 5, 2026 21:28

Merge pull request #1139 from browserstack/sdk-6463-a11y-afterEach-fa…

fbccd75

…il-guard fix(SDK-6463): catch a11y afterEach failures (cy.on fail guard) + circuit-break repeated scan timeouts

Merge branch 'master' into sdk-6463-nx-tsconfig-and-a11y-afterEach-fixes

a2fb72b

amaanbs requested changes Jul 6, 2026

View reviewed changes

chore: remove ticket identifiers from code comments

5edbea6

Addresses review feedback on #1131 — comments now describe the behavior without referencing the internal ticket. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Bhargavi-BS added a commit that referenced this pull request Jul 6, 2026

chore: remove ticket identifiers from code comments

dcac4df

Same review convention as requested on #1131. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

amaanbs approved these changes Jul 6, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(SDK-6463): robust TS config compile in NX monorepos + tolerate a11y afterEach scan timeouts#1131

fix(SDK-6463): robust TS config compile in NX monorepos + tolerate a11y afterEach scan timeouts#1131
Bhargavi-BS wants to merge 10 commits into
masterfrom
sdk-6463-nx-tsconfig-and-a11y-afterEach-fixes

Bhargavi-BS commented Jun 19, 2026

Uh oh!

amaanbs left a comment

Uh oh!

Uh oh!

Uh oh!

amaanbs left a comment

Uh oh!

Bhargavi-BS commented Jul 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Bhargavi-BS commented Jun 19, 2026

SDK-6463 — Cypress test-count / status inconsistencies (customer: Sapiens, NX monorepo)

RCA — the "inconsistency" is three layers

C.1 — TS cypress-config compile fails silently in NX monorepos → parallelization collapses

C.2 — accessibility afterEach scan/save timeout skips the rest of the spec

Validation

Files

Notes / follow-ups

Uh oh!

amaanbs left a comment

Choose a reason for hiding this comment

Automated SDK PR Review

Uh oh!

Uh oh!

Uh oh!

amaanbs left a comment

Choose a reason for hiding this comment

Uh oh!

Bhargavi-BS commented Jul 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

C.2 — accessibility `afterEach` scan/save timeout skips the rest of the spec