MySQL Bugs: #52622: valgrind warnings from rtree_page

Bug #52622	valgrind warnings from rtree_page_mgr
Submitted:	6 Apr 2010 16:15	Modified:	23 Jun 2010 12:41
Reporter:	Mark Callaghan	Email Updates:
Status:	Won't fix	Impact on me:	None
Category:	MySQL Server: GIS	Severity:	S3 (Non-critical)
Version:	5.1.45, 5.1 bzr	OS:	Any
Assigned to:	Georgi Kodinov	CPU Architecture:	Any
Tags:	gis, mtr, valgrind

Description:
I get intermittent valgrind warnings from main.gis when running mtr

==19729== Invalid write of size 8
==19729==    at 0x8782F6: rtree_page_mbr (rt_mbr.c:739)
==19729==    by 0x875014: rtree_set_key_mbr (rt_key.c:103)
==19729==    by 0x873E14: rtree_insert_level (rt_index.c:667)
==19729==    by 0x874549: rtree_insert (rt_index.c:712)
==19729==    by 0x85E18A: mi_write (mi_write.c:125)
==19729==    by 0x842862: ha_myisam::write_row(unsigned char*) (ha_myisam.cc:780)
==19729==    by 0x6AB19E: handler::ha_write_row(unsigned char*) (handler.cc:4748)
==19729==    by 0x62F33C: write_record(THD*, st_table*, st_copy_info*) (sql_insert.cc:1606)
==19729==    by 0x632B97: mysql_insert(THD*, TABLE_LIST*, List<Item>&, List<List<Item> >&, List<Item>&, List<Item>&, enum_duplicates, bool) (sql_insert.cc:835)
==19729==    by 0x5AF5D4: mysql_execute_command(THD*, unsigned long long*) (sql_parse.cc:3357)
==19729==    by 0x5B3820: mysql_parse(THD*, char const*, unsigned, char const**, unsigned long long*) (sql_parse.cc:6256)
==19729==    by 0x5B4CC7: dispatch_command(enum_server_command, THD*, char*, unsigned) (sql_parse.cc:1259)
==19729==    by 0x5B58CD: do_command(THD*) (sql_parse.cc:895)
==19729==    by 0x5A43B9: handle_one_connection (sql_connect.cc:1131)
==19729==    by 0x328F8062F6: start_thread (in /lib64/libpthread-2.5.so)
==19729==    by 0x328ECD1E3C: clone (in /lib64/libc-2.5.so)
==19729==  Address 0x58A2EC0 is just below the stack ptr.  To suppress, use: --workaround-gcc296-bugs=yes

==19729== Invalid write of size 4
==19729==    at 0x8782FD: rtree_page_mbr (rt_mbr.c:739)
==19729==    by 0x875014: rtree_set_key_mbr (rt_key.c:103)
==19729==    by 0x873E14: rtree_insert_level (rt_index.c:667)
==19729==    by 0x874549: rtree_insert (rt_index.c:712)
==19729==    by 0x85E18A: mi_write (mi_write.c:125)
==19729==    by 0x842862: ha_myisam::write_row(unsigned char*) (ha_myisam.cc:780)
==19729==    by 0x6AB19E: handler::ha_write_row(unsigned char*) (handler.cc:4748)
==19729==    by 0x62F33C: write_record(THD*, st_table*, st_copy_info*) (sql_insert.cc:1606)
==19729==    by 0x632B97: mysql_insert(THD*, TABLE_LIST*, List<Item>&, List<List<Item> >&, List<Item>&, List<Item>&, enum_duplicates, bool) (sql_insert.cc:835)
==19729==    by 0x5AF5D4: mysql_execute_command(THD*, unsigned long long*) (sql_parse.cc:3357)
==19729==    by 0x5B3820: mysql_parse(THD*, char const*, unsigned, char const**, unsigned long long*) (sql_parse.cc:6256)
==19729==    by 0x5B4CC7: dispatch_command(enum_server_command, THD*, char*, unsigned) (sql_parse.cc:1259)
==19729==    by 0x5B58CD: do_command(THD*) (sql_parse.cc:895)
==19729==    by 0x5A43B9: handle_one_connection (sql_connect.cc:1131)
==19729==    by 0x328F8062F6: start_thread (in /lib64/libpthread-2.5.so)
==19729==    by 0x328ECD1E3C: clone (in /lib64/libc-2.5.so)
==19729==  Address 0x58A2EBC is just below the stack ptr.  To suppress, use: --workaround-gcc296-bugs=yes

How to repeat:
Build mysql 5.1.45 with the facebook patch and 

./configure --enable-thread-safe-client --with-plugins=csv,blackhole,myisam,heap,innodb_plugin --without-plugin-innobase --with-fast-mutexes --with-extra-charsets=all --with-debug C_EXTRA_FLAGS="-fno-omit-frame-pointer -DHAVE_purify -DNO_ALARM -DSIGNAL_WITH_VIO_CLOSE"

centos 5.2

gcc-4.1.2

valgrind-3.2.1

Thank you for the report.

I can not repeat described behavior with current BZR sources. Does it fail without Facebook patch?

It reproduces with unmodified 5.1.45

Thank you for the feedback.

Verified as described. Problem only repeatable with gcc 4.1.2

Tried with the latest mysql-5.1-bugteam. Not able to reproduce on the following platforms:
 - sles10 32 bit gcc 4.1.2 + valgrind-3.5
 - fedora 12 64 bit gcc 4.4.3 + valgrind-3.5
 
Installing mysql sources on a fresh centos-5.2 virtual machine to try with this one as well

Wasn't able to repeat the valgrind warnings on fully yum-udated CentOS 5.2 with gcc 4.1.2 and valgrind 3.2.1 (both from the centos repositories) on the latest mysql-5.1-bugteam.

Sveta,

Can you please provide concise steps to reproduce these warnings ?

On blade12 with mysql-5.1 bzr tree:

./configure --prefix=/users/ssmirnova/blade12/build/mysql-5.1 --enable-thread-safe-client --with-plugins=csv,blackhole,myisam,heap,innodb_plugin --without-plugin-innobase --with-fast-mutexes --with-extra-charsets=all --with-debug C_EXTRA_FLAGS="-fno-omit-frame-pointer -DHAVE_purify -DNO_ALARM -DSIGNAL_WITH_VIO_CLOSE" CC=/usr/bin/gcc CXX=/usr/bin/g++

cd mysql-test

./mtr --valgrind gis

$valgrind --version
valgrind-3.2.1

Repeatable with mysql-5.1-bugteam in same environment too.

This is not repeatable in the latest pushbuild valgrind runs. I must conclude it's some oddity of the platform under test.