[debian-mysql] Bug#1053487: mariadb: FTBFS on hppa: Post-build test suite fails on multiple tests showing unexpected optimizer debug info

[Otto Kekäläinen]

Source: mariadb
Version: 1:10.11.1-1
Tags: upstream, confirmed, help, ftbfs
X-Debbugs-CC: debian-hppa at lists.debian.org
User: debian-hppa at lists.debian.org
Usertags: hppa

While reviewing the build logs from MariaDB 1:10.11.5-1 I saw multiple
tests fail with this pattern:

main.analyze_engine_stats 'innodb,slow_query_log_off' w1 [ fail ]
        Test ended at 2023-10-05 02:34:21
CURRENT_TEST: main.analyze_engine_stats
--- /<<PKGBUILDDIR>>/mysql-test/main/analyze_engine_stats.result
2023-08-11 06:35:40.000000000 +0000
+++ /<<PKGBUILDDIR>>/mysql-test/main/analyze_engine_stats.reject
2023-10-05 02:34:19.664680297 +0000
@@ -16,9 +16,6 @@
 select '$out' as X;
 X
 {
-  "query_optimization": {
-    "r_total_time_ms": "REPLACED"
-  },
   "query_block": {
     "select_id": 1,
     "r_loops": 1,
@@ -55,9 +52,6 @@
 select '$out' as X;
 X
 {
-  "query_optimization": {
-    "r_total_time_ms": "REPLACED"
-  },
   "query_block": {
     "select_id": 1,
     "r_total_time_ms": "REPLACED",
@@ -88,9 +82,6 @@
 select '$out' as X;
 X
 {
-  "query_optimization": {
-    "r_total_time_ms": "REPLACED"
-  },
   "query_block": {
     "select_id": 1,
     "r_total_time_ms": "REPLACED",
@@ -129,13 +120,9 @@
 ANALYZE FORMAT=JSON SELECT * FROM t1 WHERE a IN (SELECT s FROM t2);
 ANALYZE
 {
-  "query_optimization": {
-    "r_total_time_ms": "REPLACED"
-  },
   "query_block": {
     "select_id": 1,
     "r_loops": 1,
-    "r_total_time_ms": "REPLACED",
     "const_condition": "1",
     "nested_loop": [
       {
mysqltest: Result length mismatch

Failures were in tests: main.analyze_engine_stats,
main.derived_split_innodb, main.cte_recursive,
main.rowid_filter_innodb, main.analyze_stmt_orderby,
main.explain_json_format_partitions etc

There was also other tests failing on result mismatches to the degree
that the test suite stopped running.

Full log.
* https://buildd.debian.org/status/fetch.php?pkg=mariadb&arch=hppa&ver=1%3A10.11.5-1&stamp=1696473748&raw=0


This is not a regression in the latest MariaDB version, as there is a
similar pattern of test failues also in e.g. a log from 10.11.1:
https://buildd.debian.org/status/fetch.php?pkg=mariadb&arch=hppa&ver=1%3A10.11.1-1&stamp=1674740552&raw=0

However in the 10.11.1 build some tests passed that now failed.
Somebody with better understanding on how the optimizer debug output
works should be able to figure out what is HPPA specific here and why
it behaves differently and thus test suite ends up failing.