54 lines
3.4 KiB
Markdown
54 lines
3.4 KiB
Markdown
---
|
|
id: "bug-1964"
|
|
status: "fixed"
|
|
status_updated: "2026-10-10"
|
|
status_note: ""
|
|
component: "scheduler"
|
|
severity: "wrong-result"
|
|
timestamp: "2026-10-10T10:16:12Z"
|
|
last_seen: "2026-10-10T10:16:12Z"
|
|
occurrences: 1
|
|
files: ["src/scheduler/src/sim_limit.c", "src/scheduler/src/core_events.c"]
|
|
tags: ["scheduler", "sim", "equivalence", "limit"]
|
|
related: ["bug-1943"]
|
|
fix_commits: ["3224e906"]
|
|
gitea_issue: 395
|
|
gitea_synced_status: "fixed"
|
|
gitea_synced_hash: "97658c96f870"
|
|
---
|
|
|
|
# bug-1964: xce_scheduler --sim does not re-seed limits when a config push arrives mid-replay
|
|
|
|
## Description
|
|
**Observed:** (source reading; follow-up of the 2026-10-10 P0 fixes (bug-1936..1957))
|
|
Reported by the bug-1943 fix: the simulator seeds the limit plugin only at bootstrap (`sim_bootstrap_limit_plugin`); a CONFIG_PUSH during replay updates groups (bug-1943) but a new or changed limit config never reaches the limit plugin, while the live scheduler applies it (`plugin_config_update`).
|
|
|
|
**Expected:** Live and simulated scheduling apply the same limits after a config change.
|
|
**Impact:** Replays that span a limit config change decide differently than live.
|
|
**Found by:** follow-up of the 2026-10-10 P0 fixes (bug-1936..1957)
|
|
## Environment
|
|
- OS: no /etc/os-release; Docker container; NixOS kernel; kernel Linux 6.18.33 x86_64
|
|
- libc: glibc 2.42-47 (interpreter /nix/store/wb6rhpznjfczwlwx23zmdrrw74bayxw4-glibc-2.42-47/lib/ld-linux-x86-64.so.2)
|
|
- Toolchain: gcc (GCC) 15.2.0, meson 1.9.1
|
|
- Build: asan (xce.build)
|
|
- Code: scheduler-framework @ 19be145d-dirty
|
|
- Deployment: any (found by source reading; confirmed by the RED test of the fix)
|
|
|
|
## Reproduction
|
|
1. Replay a recording with a CONFIG_PUSH that changes a limit; compare decisions with live.
|
|
- Frequency: deterministic (by code reading; confirmed by the RED test of the fix)
|
|
- Evidence: the code sites quoted in Description; no logs kept.
|
|
## Root cause
|
|
Live `sched_init_config()` calls `plugin_config_update()`, but the core's CONFIG_PUSH handler (`core_events.c:xce_core_push_events`) did not, and the simulator loaded limits only at bootstrap (`sim_limit.c:sim_bootstrap_limit_plugin`). Pre-existing.
|
|
|
|
## Fix
|
|
`plugin_config_update_core(core, config)` (`plugin.c/.h`) is the shared update (live's `plugin_config_update()` calls it); a driver-registered core hook (`xce_core_set_config_cb`, `core.h`/`core_session.c`, modelled on the cache-clear hook, keeping the core free of plugin calls) is called by CONFIG_PUSH after loading groups; the simulator registers it; `limit_update_config()` reads `plugin->core` instead of casting `plugin->sched` (absent in sim); fixtures set `plugin->core`.
|
|
|
|
## Verification
|
|
- Test: `test_config_push_reaches_limit_plugin_in_sim_as_in_live` (`src/scheduler/testsuite/test_sim_limit_config_push_parity.c`): RED `test_sim_limit_config_push_parity.c:311: FAIL: Expected 1 Was 0 (sim: a CONFIG_PUSH during replay must configure its limit as live does)`, GREEN after; `test_sim_group_limit_parity`, `test_limit_plugin` (6/6), `test_repl_limit_table` (35/35) pass.
|
|
- Suites: `xce-build.sh all asan xce.build --incremental` on the merged branch, 2026-10-10: C suites Ok 157 / Fail 0; frontend Go tests ok (the build now also runs apiserver cmd, middleware, models and router)
|
|
- Live: n/a: unit/fixture-level verification; no live deployment was run
|
|
|
|
## References
|
|
- Fix commits: `3224e906` (on `scheduler-framework`)
|
|
- Found as a follow-up of the 2026-10-10 P0 fixes (bug-1936..1957)
|