---
id: "bug-1553"
status: "cannot-reproduce"
status_updated: "2026-09-30"
status_note: "0/26 controller_unit_test runs crashed on dc94da20, 2026-09-29; likely fixed by a later race fix (bug-1733?)"
component: "controller"
severity: "crash"
timestamp: "2026-08-13T07:35:00Z"
last_seen: "2026-09-19T09:00:00Z"
occurrences: 1
files: ["src/controller/testsuite/ (combined binary", "crash site not yet isolated -- occurs after test_drive_async_config.c's liveness_reserved_* tests)"]
tags: ["flaky", "segfault", "controller_unit_test", "order-dependent", "unity"]
related: []
repro_attempts: "0/26 runs, 2026-09-29 (20 standalone ASAN_OPTIONS=detect_leaks=0 + 6 full-suite via xce-build.sh, dc94da20, asan)"
legacy: true
---

# bug-1553: controller_unit_test (combined 1095-test Unity binary) segfaults intermittently (~1 in 3 full runs…

## Description
controller_unit_test (combined 1095-test Unity binary) segfaults intermittently (~1 in 3 full runs observed) somewhere between test_drive_async_config.c and the next testsuite file (test_configcompare.c/test_slowlane.c area) -- no assertion failure, no ASAN report captured before the crash, process just dies (exit 139/SIGSEGV)

UPDATE (2026-09-30): Not seen since August. On 2026-09-29 (dc94da20 + cfa0a941, asan) controller_unit_test ran 26 times -- 20 standalone runs (ASAN_OPTIONS=detect_leaks=0) plus 6 full-suite runs via xce-build.sh -- all exit 0, 1256/1258 Tests 0 Failures, no SIGSEGV. At the recorded ~1-in-3 rate, 20 clean standalone runs in a row has probability ~0.03%. Suspected (not proven): fixed by one of the thread-lifetime/race fixes landed since, most plausibly bug-1733 (42e1eeaa, 2026-08-29: vtask_enq wrote task->owner after publishing the task, a use-after-free that would surface as a bare SIGSEGV outside ASAN), or bug-1839/bug-1851 (controller UAFs, Sep 22/24). Proving which would need a build of an August commit reproducing ~1/3 and bisecting.

## Environment
_Not recorded (migrated from buglog.json)._

## Reproduction
_Not recorded (migrated from buglog.json)._

## Root cause
Not yet investigated. Reproduced 1/3 runs both BEFORE and AFTER an unrelated worker.c change (update_resval PERROR->rc logging fix) confirming this is pre-existing and unrelated to that fix -- likely order-dependent shared-process state leakage across Unity tests (global fixtures, xstring table, or ASAN heap state) rather than a bug in any specific test file. Two of three runs completed cleanly with 0 failures each time.

## Fix
Not fixed -- out of scope for the update_resval/mxj-auto-detect bugfix session that surfaced it. Needs isolation: bisect which specific test pair triggers it (likely by running testsuite files in smaller batches or under gdb with a core dump), then root-cause the actual memory-safety issue.

## Verification
_Not recorded (migrated from buglog.json)._

## References
_Not recorded (migrated from buglog.json)._