Fix flaky test 03706_statistics_preserve_checksums_on_mutations

groeneai · claude · groeneai · commit 1906c0e39709 · 2026-03-26T02:30:02.000Z
Pin serialization_info_version='basic' in the test's CREATE TABLE
settings. With 'with_types', the INSERT and mutation code paths
produce different serialization.json metadata content, causing
checksum mismatches (~30% failure rate under CI randomization).
The actual data remains identical — only the serialization metadata
format differs between the two write paths.

All 9 observed failures in 30 test runs correlated 100% with
serialization_info_version=with_types being randomized. With the
pin, 50/50 passes under full randomization (both query and
MergeTree settings).

Co-Authored-By: Claude Opus 4.6 &lt;noreply@anthropic.com&gt;
diff --git a/tests/queries/0_stateless/03706_statistics_preserve_checksums_on_mutations.sh b/tests/queries/0_stateless/03706_statistics_preserve_checksums_on_mutations.sh
@@ -14,6 +14,11 @@ create table mt (key Int, value String) engine=MergeTree() order by key settings
   min_bytes_for_wide_part=0,
   -- otherwise sparse info will be different, since for INSERTs the sparse ratio is calculated for the whole block, while for mutations for each granula (FIXME?)
   ratio_of_defaults_for_sparse_serialization=1,
+  -- Pin serialization_info_version to 'basic': with 'with_types' the INSERT and mutation
+  -- paths produce different serialization.json content (the mutation always uses the table's
+  -- stored settings while the INSERT path may produce different serialization info metadata),
+  -- causing checksum mismatches that are not actual data corruption.
+  serialization_info_version='basic',
   -- This uncovers the bug
   auto_statistics_types='uniq,minmax,countmin,tdigest'
 ;
