MySQL Bugs: #45882: dtoa.c might not work with gcc 4.4.0

Bug #45882	dtoa.c might not work with gcc 4.4.0
Submitted:	1 Jul 2009 14:54	Modified:	7 Jul 2010 18:52
Reporter:	Paul DuBois	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server	Severity:	S3 (Non-critical)
Version:	azalea	OS:	Any
Assigned to:	Alexey Kopytov	CPU Architecture:	Any

Description:
Per this article, gcc 4.4.0 has problems compiling dtoa.c correctly:

http://patrakov.blogspot.com/2009/03/dont-use-old-dtoac.html

We use that file (strings/dtoa.c), so we might be affected for builds that use gcc 4.4.0.

How to repeat:
See above.

Thank you for the bug report.

as a way to verify it we can try to compile with gcc 4.4.0 and run the test suite. If it'll work, perhaps SELECT "1.1"+0 (or something) will fail

The test suite fails:

main.locktrans_myisam                    [ pass ]    439
main.mysql_client_test                   [ fail ]
        Test ended at 2009-07-02 15:03:05

CURRENT_TEST: main.mysql_client_test
mysqltest: At line 14: command "$MYSQL_CLIENT_TEST --getopt-ll-test=25600M >> $MYSQLTEST_VARDIR/log/mysql_client_test.out.log 2>&1" failed

Output from before failure:
exec of '/home/miguel/dbs/azalea/bin/mysql_client_test --defaults-file=/home/miguel/dbs/azalea/mysql-test/var/my.cnf --testcase --vardir=/home/miguel/dbs/azalea/mysql-test/var --getopt-ll-test=25600M >> /home/miguel/dbs/azalea/mysql-test/var/log/mysql_client_test.out.log 2>&1' failed, error: 256, status: 1, errno: 0

The result from queries just before the failure was:
SET @old_general_log= @@global.general_log;

 - saving '/home/miguel/dbs/azalea/mysql-test/var/log/main.mysql_client_test/' to '/home/miguel/dbs/azalea/mysql-test/var/log/main.mysql_client_test/'

Only  459  of 1564 completed.
mysql-test-run: *** ERROR: Not all tests completed

however the other test suggested by Serg no:

[miguel@hegel azalea]$ bin/mysql -uroot
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1
Server version: 5.4.4-alpha-debug Source distribution

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> SELECT "1.1"+0;
+---------+
| "1.1"+0 |
+---------+
|     1.1 |
+---------+
1 row in set (0.00 sec)

mysql> exit
Bye
[miguel@hegel azalea]$ gcc --version
gcc (GCC) 4.4.0 20090506 (Red Hat 4.4.0-4)
Copyright (C) 2009 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[miguel@hegel azalea]$

Please let me know if needed more test. Thanks

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

This unit test does *not* fail in optimized build for me:
#include <gtest/gtest.h>

#include <m_string.h>

TEST(Dtoa_test, convert)
{
  char char_buf[] = "1.1";
  double d= my_atof(char_buf);
  char flt_buff[FLOATING_POINT_BUFFER];
  my_bool error= FALSE;
  size_t ret= my_fcvt(d, 1, flt_buff, &error);
  EXPECT_EQ(strlen(char_buf), ret) << flt_buff;
  EXPECT_EQ(0, strcmp(char_buf, flt_buff)) << flt_buff;
} 

But I do get lots of warnings like:
warning: dereferencing pointer ‘a.73’ does break strict-aliasing rules

We should download, and integrate
http://www.netlib.org/fp/dtoa.c

I get test failures that seem related to this bug.
I'm using next-mr-bugfixing, release build (Ver 5.6.99-m4 for Linux on x86_64) on Ubuntu 10.04 x86_64 (2.6.32-22).

Tests that fail include type_float.test, type_decimal.test, insert.test, loaddata.test and func_math.test. They all fail by hanging until they timeout.

Stacktrace for hanging type_float.test:
#0  lshift (b=0x7ffff29de580, k=3069, alloc=0x7ffff29de360)
    at /export/home/x/mysql-next-mr-bugfixing-test/strings/dtoa.c:1058
#1  0x00000000007ead75 in my_strtod_int (str=<value optimized out>, 
    end=<value optimized out>, error=<value optimized out>)
    at /export/home/x/mysql-next-mr-bugfixing-test/strings/dtoa.c:1740
#2  my_strtod (str=<value optimized out>, end=<value optimized out>, 
    error=<value optimized out>)
    at /export/home/x/mysql-next-mr-bugfixing-test/strings/dtoa.c:469
#3  0x0000000000668660 in Item_float (this=0xf37c98, str_arg=0xf37c90 "1e-50", 
    length=5) at /export/home/x/mysql-next-mr-bugfixing-test/sql/item.cc:5574
#4  0x0000000000628e13 in MYSQLparse (yythd=0xe7a530)
    at /export/home/x/mysql-next-mr-bugfixing-test/sql/sql_yacc.yy:11739

I also get this test failure in cast.test:
 SELECT f1 AS double_val, CAST(f1 AS SIGNED INT) AS cast_val FROM t1;
 double_val	cast_val
--1e30	-9223372036854775808
-1e30	9223372036854775807
-Warnings:
-Warning	1292	Truncated incorrect INTEGER value: '-1e30'
-Warning	1292	Truncated incorrect INTEGER value: '1e30'
+-1e16	-10000000000000000
+1e16	10000000000000000

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/109798

3087 Alexey Kopytov	2010-06-01
      Bug #45882: dtoa.c might not work with gcc 4.4.0
      
      - Ported relevant changes from the upstream version to not
      break strict-aliasing rules and to fix compiler warnings and
      and infinite loops caused by that issue.
      
      - Fixed compilation with Honor_FLT_ROUNDS defined.
      
      - Fixed an unused variable warning.

Pushed into 5.5.5-m3 (revid:alik@sun.com-20100615080459-smuswd9ooeywcxuc) (version source revid:marko.makela@oracle.com-20100601134335-ccthwwru23kn09qw) (merge vers: 5.1.48) (pib:16)

Pushed into mysql-next-mr (revid:alik@sun.com-20100615080558-cw01bzdqr1bdmmec) (version source revid:alik@sun.com-20100615080431-gu5icn0anrt47dsx) (pib:16)

Noted in 5.5.5 changelog.

gcc 4.4.0 could fail to compile dtoa.c.