0003722: Request for update to MySQL 5.1

(0006968)
maciej (reporter)
2009-11-10 16:19

Preliminary work is done. Current problem: tests don't pass.

Next step: make sure that MySQL builds with the debugging symbols when GARFLAVOR is set to DBG.

(0008018)
hudesd (reporter)
2010-06-14 21:26
edited on: 2010-07-01 16:38

5.1.47 is out. The Community Edition binaries from mysql for solaris work and there are performance enhancements. We want to use this for a 25 million record database.

(0008890)
maciej (reporter)
2011-03-13 12:26

I did some more work on it. It builds. I will need to re-release MySQL-5.0, and provide the binaries via the alternatives mechanism.

(0009320)
maciej (reporter)
2011-10-06 10:17

http://buildfarm.opencsw.org/experimental.html#mysql-5.1 [^]

(0009321)
hudesd (reporter)
2011-10-06 21:12

Thanks for working on 5.1. 5.5.16 is current GA (there is still support for 5.1 with the latest GA of 5.1.59). I thought I had opened a ticket for the 5.5 release but I don't see it or any note about it.

5.5 does offer some nice enhancements with performance monitoring.

I appreciate that it is not simple to do this as this affects also pm_dbd_mysql and php5_mysql (or is it php5_mysqli we're supposed to use I can't remember).

(0009322)
maciej (reporter)
2011-10-06 22:10

I looked at 5.5, and saw that it uses cmake. I had generally bad experiences with cmake when I needed to make any settings such as --prefix or --program-suffix, or additional compiler flags, or library directory changes. We'll see how it goes.

(0009323)
maciej (reporter)
2011-10-09 09:16

It generally builds, but there some things that go wrong. For example, CFLAGS are set to "-O2 -pipe -mcpu=v8", but the gcc invocations look like this:

/opt/csw/bin/gcc-4.6 -D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64 -D__STDC_
LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -O2 -pipe -mcpu=v8 -mcpu=v9 -Wall -DSIGNAL_RETURN_TYPE_IS_VOID -o CMakeFiles/cmTryCompileExec.dir/src.
c.o -c /home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv8/mysql-5.5.16/CMakeFiles/CMakeTmp/src.c

I'm positive that "-mcpu=v9" was not there in the environment flags.

(0009328)
maciej (reporter)
2011-10-10 01:23

Current compilation block: http://lists.mysql.com/packagers/517 [^]

(0009330)
maciej (reporter)
2011-10-13 09:44

I talked to guys at #cmake, and figured out that mysql-5.5.16/CMakeFiles/CMakeError.log contains the compiler invocations. Here's what I found for OpenSSL:

Building C object CMakeFiles/cmTryCompileExec.dir/CheckSymbolExists.c.o
/opt/SUNWspro/bin/cc -D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64 -D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -xO3 -m64 -xarch=sparc -KPIC -o CMakeFiles/cmTryCompileExec.dir/CheckSymbolExists.c.o -c /home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv9/mysql-5.5.16/CMakeFiles/CMakeTmp/CheckSymbolExists.c
"/home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv9/mysql-5.5.16/CMakeFiles/CMakeTmp/CheckSymbolExists.c", line 2: cannot find include file: <openssl/sha.h>

It's missing the -I/opt/csw/include parameter.

In the CMAKE invocation, these are set as environment variables, but the compiler invocation still misses the -I flag.

(0009397)
maciej (reporter)
2011-11-10 12:36

So far, mysql-5.5 has been a pain in the neck to build. Can we focus on 5.1 for the moment, and get it out the door? The binaries are here:

http://buildfarm.opencsw.org/experimental.html#mysql-5.1 [^]

(0009662)
maciej (reporter)
2012-02-26 13:26

MySQL-5.5 is available from the experimental directory: http://lists.opencsw.org/pipermail/users/2012-February/009158.html [^]

Feedback is welcome.

(0009997)
laurent (manager)
2012-07-03 17:31

5.5 from unstable branch is nice and works well for me so far (besides the unfortunate regression than needs it to be rolled down to 5.5.24).
I'm opening a separate case however because its method is apparently based on an old start script, which is inappropriate now.

(0009998)
maciej (reporter)
2012-07-03 17:45

Thanks. I'm closing this bug and we'll track the configs progress in the other bug.

Notes
(0006968) maciej (reporter) 2009-11-10 16:19	Preliminary work is done. Current problem: tests don't pass. Next step: make sure that MySQL builds with the debugging symbols when GARFLAVOR is set to DBG.

(0008018) hudesd (reporter) 2010-06-14 21:26 edited on: 2010-07-01 16:38	5.1.47 is out. The Community Edition binaries from mysql for solaris work and there are performance enhancements. We want to use this for a 25 million record database.

(0008890) maciej (reporter) 2011-03-13 12:26	I did some more work on it. It builds. I will need to re-release MySQL-5.0, and provide the binaries via the alternatives mechanism.

(0009320) maciej (reporter) 2011-10-06 10:17	http://buildfarm.opencsw.org/experimental.html#mysql-5.1 [^]

(0009321) hudesd (reporter) 2011-10-06 21:12	Thanks for working on 5.1. 5.5.16 is current GA (there is still support for 5.1 with the latest GA of 5.1.59). I thought I had opened a ticket for the 5.5 release but I don't see it or any note about it. 5.5 does offer some nice enhancements with performance monitoring. I appreciate that it is not simple to do this as this affects also pm_dbd_mysql and php5_mysql (or is it php5_mysqli we're supposed to use I can't remember).

(0009322) maciej (reporter) 2011-10-06 22:10	I looked at 5.5, and saw that it uses cmake. I had generally bad experiences with cmake when I needed to make any settings such as --prefix or --program-suffix, or additional compiler flags, or library directory changes. We'll see how it goes.

(0009323) maciej (reporter) 2011-10-09 09:16	It generally builds, but there some things that go wrong. For example, CFLAGS are set to "-O2 -pipe -mcpu=v8", but the gcc invocations look like this: /opt/csw/bin/gcc-4.6 -D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64 -D__STDC_ LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -O2 -pipe -mcpu=v8 -mcpu=v9 -Wall -DSIGNAL_RETURN_TYPE_IS_VOID -o CMakeFiles/cmTryCompileExec.dir/src. c.o -c /home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv8/mysql-5.5.16/CMakeFiles/CMakeTmp/src.c I'm positive that "-mcpu=v9" was not there in the environment flags.

(0009328) maciej (reporter) 2011-10-10 01:23	Current compilation block: http://lists.mysql.com/packagers/517 [^]

(0009330) maciej (reporter) 2011-10-13 09:44	I talked to guys at #cmake, and figured out that mysql-5.5.16/CMakeFiles/CMakeError.log contains the compiler invocations. Here's what I found for OpenSSL: Building C object CMakeFiles/cmTryCompileExec.dir/CheckSymbolExists.c.o /opt/SUNWspro/bin/cc -D_LARGEFILE_SOURCE -D_LARGE_FILES -D_FILE_OFFSET_BITS=64 -D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -xO3 -m64 -xarch=sparc -KPIC -o CMakeFiles/cmTryCompileExec.dir/CheckSymbolExists.c.o -c /home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv9/mysql-5.5.16/CMakeFiles/CMakeTmp/CheckSymbolExists.c "/home/maciej/src/opencsw/pkg/mysql5/branches/mysql-5.5.x/work/solaris9-sparc/build-isa-sparcv9/mysql-5.5.16/CMakeFiles/CMakeTmp/CheckSymbolExists.c", line 2: cannot find include file: <openssl/sha.h> It's missing the -I/opt/csw/include parameter. In the CMAKE invocation, these are set as environment variables, but the compiler invocation still misses the -I flag.

(0009397) maciej (reporter) 2011-11-10 12:36	So far, mysql-5.5 has been a pain in the neck to build. Can we focus on 5.1 for the moment, and get it out the door? The binaries are here: http://buildfarm.opencsw.org/experimental.html#mysql-5.1 [^]

(0009662) maciej (reporter) 2012-02-26 13:26	MySQL-5.5 is available from the experimental directory: http://lists.opencsw.org/pipermail/users/2012-February/009158.html [^] Feedback is welcome.

(0009997) laurent (manager) 2012-07-03 17:31	5.5 from unstable branch is nice and works well for me so far (besides the unfortunate regression than needs it to be rolled down to 5.5.24). I'm opening a separate case however because its method is apparently based on an old start script, which is inappropriate now.

(0009998) maciej (reporter) 2012-07-03 17:45	Thanks. I'm closing this bug and we'll track the configs progress in the other bug.

OpenCSW Bug Tracker