worked around gcc bug to make compile-pentium-gcov work

increase coverage for table.cc in mysql_rm_table()
ingore temporary gcov files
fixed bug in replication test case


BitKeeper/etc/ignore:
  added *.da
BUILD/compile-pentium-gcov:
  do not build shared libraries - combination of -fPIC and the coverage flags
  triggers a bug in gcc, apparently - problem went away after adding
  --disable-shared
mysql-test/t/flush.test:
  extra test to increase coverage of table.cc
mysql-test/t/rpl000016.test:
  fixed bug in test case - insert may be concurrent, so we may read the
  table on the slave before the last insert becomes visible even when the 
  master position is up to date if we do not
  lock it explicitly

.bzrignore

@@ -199,3 +199,7 @@ bdb/dist/tags
 bdb/build_unix/*
 sql/.gdbinit
 BitKeeper/etc/level
+*.gcov
+*.bb
+*.bbg
+*.da

BUILD/compile-pentium-gcov

@@ -4,6 +4,6 @@ path=`dirname $0`
 . "$path/SETUP.sh"
 
 extra_flags="$pentium_cflags -O2 -fprofile-arcs -ftest-coverage"
-extra_configs="$pentium_configs $debug_configs"
+extra_configs="$pentium_configs $debug_configs --disable-shared"
 
 . "$path/FINISH.sh"

mysql-test/t/flush.test

@@ -22,4 +22,13 @@ while ($1)
 
 connection con1;
 select * from t1;
+connection con2;
+flush tables with read lock;
+--error 1099;
 drop table t2;
+connection con1;
+send drop table t2;
+connection con2;
+unlock tables;
+connection con1;
+reap;

mysql-test/t/rpl000016.test

@@ -85,7 +85,11 @@ slave stop;
 slave start;
 sync_with_master;
 show slave status;
+# because of concurrent insert, the table may not be up to date
+# if we do not lock
+lock tables t3 read;
 select count(*) from t3 where n = 4;
+unlock tables;
 #clean up
 connection master;
 drop table if exists t1,t2,t3;