RUN_UNIT_TESTS_SEQUENTIAL is useful to detect cases where global state is left dirty in the test process and leads to subsequent unit test cases failing. However, one CI task is sufficient to catch this.
As there already is one, add docs there and remove this env var (and extra logic).
The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.
For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/33136.
See the guideline for information on the review process. A summary of reviews will appear here.
No conflicts as of last run.
373@@ -374,7 +374,7 @@ jobs:
374 - name: Run unit tests
375 # Can't use ctest here like other jobs as we don't have a CMake build tree.
376 run: |
377- ./bin/test_bitcoin.exe -l test_suite
378+ ./bin/test_bitcoin.exe -l test_suite # Intentionally run sequentially here, to catch test case failures caused by dirty global state from prior test cases.
to catch test case failures caused by dirty global state from prior test cases.
If the aim is to catch dirty state, should we run it in a job with the most unit tests? Windows cross-build is going to run less than Linux.
I think it is always possible to miss unit tests that are not compiled. For the Windows ones, it should only be a few about windows paths, fork(), and socketpair, which seems fine to skip.
If there is need to run it in another task, it can be done as a follow-up?