Let's see if these changes will persuade the regression tests to pass in

koji.

mysql-no-check-testcases.patch

@@ -0,0 +1,20 @@
+As of mysql 5.1.x, the --check-testcases option in mysql-test-run was made
+the default, although the documentation for it doesn't tell you that.
+Unfortunately the check-testcases code frequently fails in koji for
+obscure reasons; what's more it slows down the tests materially.
+Since it's not really our job to debug the test suite, let's just turn
+it back off ...
+
+
+diff -Naur mysql-5.1.37.orig/mysql-test/mysql-test-run.pl mysql-5.1.37/mysql-test/mysql-test-run.pl
+--- mysql-5.1.37.orig/mysql-test/mysql-test-run.pl	2009-07-13 19:09:03.000000000 -0400
++++ mysql-5.1.37/mysql-test/mysql-test-run.pl	2009-08-14 17:40:49.000000000 -0400
+@@ -193,7 +193,7 @@
+ 
+ my $opt_skip_core;
+ 
+-our $opt_check_testcases= 1;
++our $opt_check_testcases= 0;
+ my $opt_mark_progress;
+ 
+ my $opt_sleep;

mysql-testing.patch

@@ -1,6 +1,13 @@
 Hack the top-level Makefile to enable the openssl regression tests.
 (Why doesn't this happen automatically given the configure option??)
 
+Also, increase the overall timeout for the regression tests to 12 hours,
+because on a slow or heavily-loaded build machine sometimes the default of
+5 hours isn't enough.  (This has been demonstrated to fail in mass-rebuild
+scenarios, which aren't that uncommon for Fedora.)  Similarly increase the
+per-testcase timeout to 30 minutes, since the default of 15 hasn't got a
+great deal of headroom either.
+
 
 diff -Naur mysql-5.1.32.orig/Makefile.am mysql-5.1.32/Makefile.am
 --- mysql-5.1.32.orig/Makefile.am	2009-02-13 19:51:56.000000000 -0500
@@ -10,7 +17,7 @@ diff -Naur mysql-5.1.32.orig/Makefile.am mysql-5.1.32/Makefile.am
  test-ns:
  	cd mysql-test ; \
 -	    @PERL@ ./mysql-test-run.pl $(force) $(mem) --mysqld=--binlog-format=mixed
-+	    @PERL@ ./mysql-test-run.pl $(force) $(mem) --ssl --mysqld=--binlog-format=mixed
++	    @PERL@ ./mysql-test-run.pl $(force) $(mem) --ssl --mysqld=--binlog-format=mixed --suite-timeout=720 --testcase-timeout=30
  
  test-binlog-statement:
  	cd mysql-test ; \

mysql.spec

@@ -1,6 +1,6 @@
 Name: mysql
 Version: 5.1.37
-Release: 1%{?dist}
+Release: 2%{?dist}
 Summary: MySQL client programs and shared libraries
 Group: Applications/Databases
 URL: http://www.mysql.com
@@ -32,6 +32,7 @@ Patch6: mysql-stack-guard.patch
 Patch7: mysql-plugin-bug.patch
 Patch8: mysql-setschedparam.patch
 Patch9: mysql-ndb-stacksize.patch
+Patch10: mysql-no-check-testcases.patch
 
 BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-root
 BuildRequires: gperf, perl, readline-devel, openssl-devel
@@ -174,6 +175,7 @@ the MySQL sources.
 %patch7 -p1
 %patch8 -p1
 %patch9 -p1
+%patch10 -p1
 
 libtoolize --force
 aclocal
@@ -249,6 +251,7 @@ make check
 
   # note: "make test" invokes two largely-duplicate sets of tests,
   # which makes the runtime really unacceptably long ...
+  # if you want to change this, look at mysql-testing.patch too.
   make test-ns
 %endif
 
@@ -639,6 +642,10 @@ fi
 %{_mandir}/man1/mysql_client_test.1*
 
 %changelog
+* Fri Aug 14 2009 Tom Lane <tgl@redhat.com> 5.1.37-2
+- Add a couple of patches to improve the probability of the regression tests
+  completing in koji builds
+
 * Sun Aug  2 2009 Tom Lane <tgl@redhat.com> 5.1.37-1
 - Update to MySQL 5.1.37, for various fixes described at
   http://dev.mysql.com/doc/refman/5.1/en/news-5-1-37.html