MySQL Bugs: #47146: Linking problem with example plugin when dtrace enabled

Bug #47146	Linking problem with example plugin when dtrace enabled
Submitted:	4 Sep 2009 19:44	Modified:	4 Aug 2010 22:54
Reporter:	Bjørn Munch	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: General	Severity:	S3 (Non-critical)
Version:	5.4	OS:	Solaris
Assigned to:	Horst Hunger	CPU Architecture:	Any
Tags:	DTrace, experimental, plugin

Description:
Tests main.plugin and rpl.rpl_plugin_load fail on Solaris in -trunk where dtrace is enabled by default in configure:

mysqltest: At line 7: query 'INSTALL PLUGIN example SONAME $HA_EXAMPLE_SO' failed: 1126: Can't open shared library '/tmp/t/storage/example/.libs/ha_example.so' (errno: 2 ld.so.1: mysqld: fatal: relocation error: file /tmp/t/storage/example/.libs/ha_example.so: symbol __dtrace_mysql___index__read__)

(Part of the symbol name has been cut off)

How to repeat:
Run either of these tests on Solaris when built with dtrace, also seen in PB2.

It has not been seen before in PB2 because those tests have been incorrectly skipped, but after Bug #45298 was fixed, they surface.  I will set them to 'experimental' on Solaris in mysql-trunk-mtr before merging this to mysql-trunk.

Horst, you are our Dtrace/plugin guy :-) So please take a look at this. Thank you.

Would you, please, show me the configure call with parameters?

Settings ang config line from PB2 where this bug is seen. I cannot add it as comment since the lines get wrapped.

Attachment: conf (application/octet-stream, text), 1.21 KiB.

It seems so, as if the example plugin will not be generated to be dynamically linkable, see configure parameter for plugins.

[Bug #42144 (which has the symptoms I am experiencing when building 5.1 on Solaris 10/SPARC), refers over to this bug.]

Can someone confirm whether this is just a test-case bug, or whether I need to disable DTrace support in order to build a clean MySQL?

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/85044

2880 Horst Hunger	2009-09-29
      Patch for Bug#47146. The dtrace was never executed due to wrong dependencies and the wrong .o was used when linking the shared library. Now dtrace will be executed and .libs/ha_example_la-ha_example.o will be used in the shared library.

Wrote a review and mailed it.

To me, it doesn't yet look correct, especially when comparing with how it is done for other handlers.
Maybe they got it wrong, too - this should be checked.

In any case, the patch has some inconsistencies which should be fixed.

Some tests in mysql-test/collections/default.experimantal have been tagged with this bug number, please remove these as part of any fix.

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/88159

2915 Horst.Hunger	2009-10-26
      Patch for bug#47146: Patch after first review. Taken from the Makefile.am for archive and changed for example.

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/88313

2915 Horst.Hunger	2009-10-27
      Patch for bug#47146: Patch after first review. Taken from the Makefile.am for archive and changed for example. Copyright line has been changed.

I commented that the entries in mysql-test/collections/default.experimantal that were tagged with this test should be removed, but the patch does not do so. Please fix that before pushing.

The 'dtrace_shared_files' target is not required, it's not used anywhere. Please remove it.

The same applies to the patch for bug #47739, I had overlooked that when reviewing it, sorry about that.

I do not agree to Alexey's comment:
AIUI, neither "dtrace_files" nor "dtrace_shared_files" nor "dtrace_providers" nor "dtrace_sources" are required technically, so they might all be removed.

However, I think they serve documentation purposes during the build, so I really see merit in their existence - provided they all get removed during cleanup.

Because of this, I would like to see "dtrace_sources" added to "CLEANFILES".
With this addition and Horst's tests that it works, I approve of the change.

I added 'dtrace_files' targets because sql/Makefile.am needs to process all object files in one step, so it has to know names of all object files containing probes from all libraries (or storage engines). 

Similarly, 'dtrace_providers' targets are required by sql/Makefile.am to track all providers defined by libraries/storage engines, since storage engines will not necessarily put their probes into the 'mysql' provider, they are free to define their own ones.

And we need to track all object files and providers to work around a well know DTrace limitation in processing static archives. See sql/Makefile.am for details. 

Shared libraries are not affected by that problem, we don't need to track object files containing probes that are linked into a shared library to process them in sql/Makefile.am. Therefore, we don't need the 'dtrace_shared_files' targets.

'dtrace_sources' wasn't added by me. It is not actually used anywhere and its purpose eludes me. It seems like someone just copy-pasted the code for 'link_sources' in client/Makefile.am.

My conclusion is: Remove "dtrace_shared_files" as target as I completely agree with Alexeys explanation. Alexey agrees in considering the review as done.

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/89488

2915 Horst.Hunger	2009-11-05
      Patch for bug#47146: Patch after review. Taken from the Makefile.am for archive and changed for example. Copyright line has been changed. The target "dtrace_shared_file" has been removed.

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/89494

2940 Horst.Hunger	2009-11-05
      Patch for bug#47146: Patch after review. Taken from the Makefile.am for archive and changed for example. Copyright line has been changed. The target "dtrace_shared_file" has been removed.

pushed to mysql-next-mr-bugfixing.

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/90069

2945 Philip Stoev	2009-11-11
      revert patch for bug #47146 because it breaks Solaris builds

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/90071

3703 Philip Stoev	2009-11-11 [merge]
      merge: revert patch for bug #47146 because it breaks Solaris builds

Patch for Bug 47146

Attachment: 47146.patch (application/octet-stream, text), 1001 bytes.

Pushed into 6.0.14-alpha (revid:kostja@sun.com-20091120124947-yi6h2jbgw0kbciwm) (version source revid:pstoev@mysql.com-20091111100316-9dumrv0rxrl8g67j) (merge vers: 6.0.14-alpha) (pib:13)

Since the patch committed by the assignee was reverted, I am setting status back to Verified. MC has suggested/attached an alternative patch, but it does not look like it has been committed yet, so status Documenting would be wrong.

Pushed into 5.6.0-beta (revid:alik@sun.com-20091211055628-ltr7fero363uev7r) (version source revid:alik@sun.com-20091211055453-717czhtezc74u8db) (merge vers: 5.6.0-beta) (pib:13)

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/94475

2934 Horst.Hunger	2009-12-16
      This is a the same patch that MC committed ago to bug#47146. I need it for a further push. This patch has been tested with cc and gcc on Solaris.

See also Bug#42144.

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/102150

2985 Horst.Hunger	2010-03-03
      Bug#47146 : Patch based on "mysql-next-mr-bugfixing". Includes MC's patch and the changes of the first review.

Tested patch on Solaris 10 on AMD and SPARC using Sun Studio compiler. Comments sent by email to the commits list.

Summary: it is necessary to produce position independent code (PIC) for the storage engine plug-ins. To ensure this is done when using the Sun Studio compiler the -Kpic or -KPIC option must be given to the compiler.

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/102462

2985 Horst.Hunger	2010-03-05
      Bug#47146 : Patch based on "mysql-next-mr-bugfixing". Includes MC's patch and the changes required by the reviews.

Pushed into 5.5.3-m3 (revid:alik@sun.com-20100306103849-hha31z2enhh7jwt3) (version source revid:vvaintroub@mysql.com-20091211201717-03qf8ckwiw0np80p) (merge vers: 5.6.0-beta) (pib:16)

Bug#52211 has been marked as a duplicate of this bug.

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/104124

2992 Alexey Kopytov	2010-03-23
      Mark all main.plugin* tests as experimental on Solaris
      as done in 6.0 due to bug #47146.

Pushed into 5.5.4-m3 (revid:alik@sun.com-20100326080914-2pz8ns984e0spu03) (version source revid:alexey.kopytov@sun.com-20100323203412-xr3vul0peew7jioc) (merge vers: 5.5.3-m2) (pib:16)

Pushed into mysql-next-mr (revid:alik@sun.com-20100326081116-m3v4l34yhr43mtsv) (version source revid:alik@sun.com-20100325072612-4sds00ix8ajo1e84) (pib:16)

Pushed into 6.0.14-alpha (revid:alik@sun.com-20100326081944-qja07qklw1p2w7jb) (version source revid:alik@sun.com-20100325090207-0462f2l8qhgy9qfa) (merge vers: 6.0.14-alpha) (pib:16)

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/106244

3145 Horst.Hunger	2010-04-21
      This patch for bug#47146 needs the patch "http://lists.mysql.com/commits/102462" from 2010-03-05 as base.

Pushed the last patch to next-mr-bugfixing and 6.0-codebase-bugfixing.

rpl_ndb_binlog_format_errors fails in mysql-trunk
on OpenSolaris due to this bug.

Should not the patch be pushed into mysql-trunk-bugfixing?

Pushed into 6.0.14-alpha (revid:alik@sun.com-20100427094135-5s49ecp3ckson6e2) (version source revid:alik@sun.com-20100427093843-uekr85qkd7orx12t) (merge vers: 6.0.14-alpha) (pib:16)

Pushed into mysql-next-mr (revid:alik@sun.com-20100427094036-38frbg3famdlvjup) (version source revid:alik@sun.com-20100427093825-92wc8b22d4yg34ju) (pib:16)

A note has been added to the 5.5.4, 6.0.14 changelogs: 

        When using the <literal>EXAMPLE</literal> storage engine, when                                                                                     
        the engine had been built as a plugin (instead of built-in), and                                                                                   
        DTrace probes had been enabled during the build, loading the                                                                                       
        storage engine library would fail due to a missing object table                                                                                    
        entry.

Pushed into mysql-trunk 5.6.1-m4 (revid:alik@ibmvm-20100804080001-bny5271e65xo34ig) (version source revid:alik@sun.com-20100427095914-pzlxbqjjtnngxmf0) (merge vers: 5.6.99-m4) (pib:18)

Pushed into mysql-trunk 5.6.1-m4 (revid:alik@ibmvm-20100804081533-c1d3rbipo9e8rt1s) (version source revid:alik@sun.com-20100427095914-pzlxbqjjtnngxmf0) (merge vers: 5.6.99-m4) (pib:18)

Bug does not appear in any released 5.6.x version.