_run_integration_tests() captures result.returncode from the pytest subprocess but no caller (complete_module, _complete_module_quiet, run_module_tests) ever reads it -- only 'passed'/'failed' are checked.
If pytest fails during collection (e.g. an import-time error in the exported package -- distinct from the notebook-syntax check, which only compile()s notebook cells, not the exported package), it exits nonzero with no PASSED/FAILED lines to parse. _parse_pytest_output returns an empty list, so the result comes back as passed=0, failed=0.
Every caller's if integration_result['failed'] > 0 check is then False, so tito module complete prints "No integration tests for this module" and marks the module complete despite the exported package being unimportable/broken.
Fix
Treat "zero parsed tests + nonzero pytest exit code" as a collection failure rather than "no tests exist", surfacing it through the same tests-list/error-field shape _parse_test_output already uses for its own unit-test fallback case. The legitimate "no test file for this module" short-circuit (returns before pytest even runs) and the "pytest ran fine but collected zero tests" case (returncode == 0) are unaffected.
Test plan
Fed _parse_pytest_output a real pytest collection-error transcript (ImportError during test collection): confirmed it returns zero parsed tests, matching the bug's premise.
Confirmed the new guard turns that into failed=1 instead of the old passed=0/failed=0, which now correctly trips the existing if integration_result['failed'] > 0 branch in every caller.
python -m py_compile passes on the edited file.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.
## 📋 Pull Request Information
**Original PR:** https://github.com/harvard-edge/cs249r_book/pull/1951
**Author:** [@Shashank-Tripathi-07](https://github.com/Shashank-Tripathi-07)
**Created:** 7/7/2026
**Status:** 🔄 Open
**Base:** `dev` ← **Head:** `fix/workflow-integration-test-collection-failure`
---
### 📝 Commits (1)
- [`77ba36f`](https://github.com/harvard-edge/cs249r_book/commit/77ba36ff399f2b19670babeddf5078fb08bffd85) fix(tinytorch): integration test collection failure silently treated as no-tests
### 📊 Changes
**1 file changed** (+13 additions, -0 deletions)
<details>
<summary>View changed files</summary>
📝 `tinytorch/tito/commands/module/workflow.py` (+13 -0)
</details>
### 📄 Description
## Summary
- `_run_integration_tests()` captures `result.returncode` from the pytest subprocess but no caller (`complete_module`, `_complete_module_quiet`, `run_module_tests`) ever reads it -- only `'passed'`/`'failed'` are checked.
- If pytest fails during **collection** (e.g. an import-time error in the exported package -- distinct from the notebook-syntax check, which only `compile()`s notebook cells, not the exported package), it exits nonzero with no `PASSED`/`FAILED` lines to parse. `_parse_pytest_output` returns an empty list, so the result comes back as `passed=0, failed=0`.
- Every caller's `if integration_result['failed'] > 0` check is then `False`, so `tito module complete` prints `"No integration tests for this module"` and marks the module complete despite the exported package being unimportable/broken.
## Fix
Treat "zero parsed tests + nonzero pytest exit code" as a collection failure rather than "no tests exist", surfacing it through the same tests-list/error-field shape `_parse_test_output` already uses for its own unit-test fallback case. The legitimate "no test file for this module" short-circuit (returns before pytest even runs) and the "pytest ran fine but collected zero tests" case (`returncode == 0`) are unaffected.
## Test plan
- [x] Fed `_parse_pytest_output` a real pytest collection-error transcript (`ImportError` during test collection): confirmed it returns zero parsed tests, matching the bug's premise.
- [x] Confirmed the new guard turns that into `failed=1` instead of the old `passed=0/failed=0`, which now correctly trips the existing `if integration_result['failed'] > 0` branch in every caller.
- [x] `python -m py_compile` passes on the edited file.
---
<sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
📋 Pull Request Information
Original PR: https://github.com/harvard-edge/cs249r_book/pull/1951
Author: @Shashank-Tripathi-07
Created: 7/7/2026
Status: 🔄 Open
Base:
dev← Head:fix/workflow-integration-test-collection-failure📝 Commits (1)
77ba36ffix(tinytorch): integration test collection failure silently treated as no-tests📊 Changes
1 file changed (+13 additions, -0 deletions)
View changed files
📝
tinytorch/tito/commands/module/workflow.py(+13 -0)📄 Description
Summary
_run_integration_tests()capturesresult.returncodefrom the pytest subprocess but no caller (complete_module,_complete_module_quiet,run_module_tests) ever reads it -- only'passed'/'failed'are checked.compile()s notebook cells, not the exported package), it exits nonzero with noPASSED/FAILEDlines to parse._parse_pytest_outputreturns an empty list, so the result comes back aspassed=0, failed=0.if integration_result['failed'] > 0check is thenFalse, sotito module completeprints"No integration tests for this module"and marks the module complete despite the exported package being unimportable/broken.Fix
Treat "zero parsed tests + nonzero pytest exit code" as a collection failure rather than "no tests exist", surfacing it through the same tests-list/error-field shape
_parse_test_outputalready uses for its own unit-test fallback case. The legitimate "no test file for this module" short-circuit (returns before pytest even runs) and the "pytest ran fine but collected zero tests" case (returncode == 0) are unaffected.Test plan
_parse_pytest_outputa real pytest collection-error transcript (ImportErrorduring test collection): confirmed it returns zero parsed tests, matching the bug's premise.failed=1instead of the oldpassed=0/failed=0, which now correctly trips the existingif integration_result['failed'] > 0branch in every caller.python -m py_compilepasses on the edited file.🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.