From 6147bf61869499552ce806ecc994bf9fd0151983 Mon Sep 17 00:00:00 2001 From: Josh Date: Thu, 15 Oct 2026 21:34:16 +0200 Subject: [PATCH] test(ibd): serialise h-slow GetData latch reset with rotate-once tests r219_h_slow_skips_dump_rotates_fat_once cleared process-wide LAST_H_SLOW_GD_MS without serial(ibd), so it could wipe the latch while r219_h_slow_dump_keeps_preferred_fat_rotates_once expected the same GetData sample not to rotate again. --- src/node/parallel_ibd/chunk_assigner_tests.rs | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/node/parallel_ibd/chunk_assigner_tests.rs b/src/node/parallel_ibd/chunk_assigner_tests.rs index 330b387..0a11cb5 100644 --- a/src/node/parallel_ibd/chunk_assigner_tests.rs +++ b/src/node/parallel_ibd/chunk_assigner_tests.rs @@ -14518,6 +14518,9 @@ fn r313_no_farm_default_off() { } /// R-218 dump hole must not rotate. Fat slow sample rotates once. +/// `test_reset_h_slow_gd` clears process-wide `LAST_H_SLOW_GD_MS`; keep it on +/// the same `serial(ibd)` lane as the rotate-once tests that rely on that latch. +#[serial_test::serial(ibd)] #[test] fn r219_h_slow_skips_dump_rotates_fat_once() { super::test_reset_h_slow_gd();