Regressions in code coverage in sql_join_cache in maria-5.3-mwl128

[Philip Stoev]

The code coverage for sql_join_cache.cc has gone down significantly in maria-5.3-mwl128 as compared to 5.3-main. The report is available here:

http://fedora13.selfip.org/lcov/maria-5.3/sql/sql_join_cache.cc.gcov.html
http://fedora13.selfip.org/lcov/maria-5.3-mwl128/sql/sql_join_cache.cc.gcov.html

The following items are not covered:

* the bottom of JOIN_CACHE::alloc_buffer()

* Several JOIN_CACHE_HASHED and JOIN_CACHE_BNLH functions

* several bkah* and JOIN_CACHE_BKAH functions

[Philip Stoev (pstoev-askmonty)]

The reason seems to be that the join_cache test failed in the gcov build:

main.join_cache w1 [ fail ]
        Test ended at 2010-10-23 19:22:36

CURRENT_TEST: main.join_cache
--- /home/buildbot/rqg-buildbot-slave/perpush-gcov/build/mysql-test/r/join_cache.result 2010-10-23 18:43:27.124943000 +0300
+++ /home/buildbot/rqg-buildbot-slave/perpush-gcov/build/mysql-test/r/join_cache.reject 2010-10-23 19:22:36.159247624 +0300
@@ -4041,8 +4041,8 @@
 SГЈo Paulo Brazil
 London United Kingdom
 Santiago de Chile Chile
-Alexandria Egypt
 Cairo Egypt
+Alexandria Egypt
 Jakarta Indonesia
 Delhi India
 Calcutta [Kolkata] India

...

Build and compile information are available from http://buildbot.askmonty.org/buildbot/builders/gcov/builds/38 . No idea why.

[Philip Stoev (pstoev-askmonty)]

Igor pushed a --sort-result test case fix and the coverage percentages are back to normal.