ci: nightly Valgrind memcheck failed #1005

Activity

1. github-actions

   added

   bugSomething isn't workingSomething isn't working

   area: ciSubsystem: ciSubsystem: ci

   nightly-check:valgrindAuto-filed by nightly-slow-checks.yml for the Valgrind legAuto-filed by nightly-slow-checks.yml for the Valgrind leg

   on Oct 20, 2026
2. github-actions commented on Oct 21, 2026

   

   github-actionsBot

   on Oct 21, 2026 – with GitHub Actions

   ContributorAuthor

   More actions

   Scheduled nightly check failed: Valgrind memcheck.

   Verification status

   Reproduced. This is a real scheduled CI run, not inferred from
   reading code. Revision: 16ed13de24f30c2b3c7243f309675d0bf027caac. Run: https://github.com/LASTRADA-Software/morph/actions/runs/39570393171

   Real output

   Job: Valgrind memcheck (nightly)
   Run: https://github.com/LASTRADA-Software/morph/actions/runs/39570393171
   Commit: 16ed13de24f30c2b3c7243f309675d0bf027caac
   

   What would change the verdict

   This issue closes automatically the next time this job's scheduled
   run passes (see .github/workflows/nightly-slow-checks.yml's
   reconcile-issues job, matched by the nightly-check:valgrind label
   below). Re-opens itself in the same way if it fails again after
   that.
3. github-actions commented on Oct 22, 2026

   

   github-actionsBot

   on Oct 22, 2026 – with GitHub Actions

   ContributorAuthor

   More actions

   Scheduled nightly check failed: Valgrind memcheck.

   Verification status

   Reproduced. This is a real scheduled CI run, not inferred from
   reading code. Revision: 16ed13de24f30c2b3c7243f309675d0bf027caac. Run: https://github.com/LASTRADA-Software/morph/actions/runs/39749154902

   Real output

   Job: Valgrind memcheck (nightly)
   Run: https://github.com/LASTRADA-Software/morph/actions/runs/39749154902
   Commit: 16ed13de24f30c2b3c7243f309675d0bf027caac
   

   What would change the verdict

   This issue closes automatically the next time this job's scheduled
   run passes (see .github/workflows/nightly-slow-checks.yml's
   reconcile-issues job, matched by the nightly-check:valgrind label
   below). Re-opens itself in the same way if it fails again after
   that.
4. github-actions commented on Oct 23, 2026

   

   github-actionsBot

   on Oct 23, 2026 – with GitHub Actions

   ContributorAuthor

   More actions

   Scheduled nightly check failed: Valgrind memcheck.

   Verification status

   Reproduced. This is a real scheduled CI run, not inferred from
   reading code. Revision: f13c29a97cdcb065f74ea3fd364eab800f91aa04. Run: https://github.com/LASTRADA-Software/morph/actions/runs/39925492763

   Real output

   Job: Valgrind memcheck (nightly)
   Run: https://github.com/LASTRADA-Software/morph/actions/runs/39925492763
   Commit: f13c29a97cdcb065f74ea3fd364eab800f91aa04
   

   What would change the verdict

   This issue closes automatically the next time this job's scheduled
   run passes (see .github/workflows/nightly-slow-checks.yml's
   reconcile-issues job, matched by the nightly-check:valgrind label
   below). Re-opens itself in the same way if it fails again after
   that.
5. Yaraslaut commented on Oct 23, 2026

   

   Yaraslaut

   on Oct 23, 2026

   Member

   More actions

   Triage: memcheck finds no memory errors; one test fails under Valgrind, so this is a timing problem

   Measured from the job logs of the last four nightly runs (39394274700, 39570393171, 39749154902, 39925492763). Not reproduced locally: morph_net_tests isn't built in a local Valgrind configuration yet.

   In last night's run on f13c29a97, memcheck reported no errors in any of the three suites:

   ── memcheck: tests/morph_tests ──
   test cases:  3136 |  3135 passed | 1 failed as expected
   ==7924== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
   ── memcheck: tests/net/morph_net_tests ──
   test cases:  258 |  257 passed | 1 failed
   ==10385== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
   ── memcheck: tests/offline_sqlite/morph_offline_sqlite_tests ──
   ==12468== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
   

   The job fails because one test case fails under Valgrind:

   tests/net/test_push_net.cpp:286  net_connect_handler_fires_on_first_and_reconnect
   tests/net/test_push_net.cpp:324: FAILED:
     REQUIRE( ::morph::testing::waitUntil([&] { return snapshot().size() == 1; }, budget) )
   

   test_push_net.cpp also failed on 39394274700 and 39749154902, at :322 before #1037 moved the line. #1037 raised this wait to a 10 s WaitBudget, and it still fails on f13c29a97. So either 10 s is still too short under Valgrind on a shared runner, or the first connect never happens in that process. One candidate, not verified: the earlier RLIMIT_NOFILE tests in test_socket_server.cpp make Valgrind print invalid file descriptor … in syscall socket() at the start of the suite. The run was also previously red for other reasons (replay_ledger_conformance.hpp:52, the ui use-after-free fixed in #1037), so last night is the first run where this test is the only cause.

   The assert aborts in registry.hpp/bridge.hpp higher up the log are expected: test_registration_phase.cpp runs them in forked children on purpose, and each child reports ERROR SUMMARY: 0 errors.

   Verdict: rescope. No defect in include/morph or src is shown. This is a test timing or isolation problem under Valgrind, which per AGENTS.md is fixed in place, not tracked as a library bug.

   Would change the verdict: a nonzero memcheck ERROR SUMMARY, or the push test failing outside Valgrind (it passes in every other leg).
6. Yaraslaut

   added

   triage: rescopeReal problem, wrong framing; rewrite before buildingReal problem, wrong framing; rewrite before building

   and removed

   triage: rescopeReal problem, wrong framing; rewrite before buildingReal problem, wrong framing; rewrite before building

   on Oct 23, 2026
7. Yaraslaut commented on Oct 23, 2026

   

   Yaraslaut

   on Oct 23, 2026

   Member

   More actions

   Closing: memcheck reports no memory errors in any suite, so nothing in include/morph or src needs fixing. What fails is a test-timing problem under Valgrind (test_push_net.cpp:324), and per AGENTS.md that gets fixed in place rather than tracked. The nightly reconcile job reopens this issue if the job fails again.
8. Yaraslaut

   closed this as not plannedon Oct 23, 2026