Bug #36124 Using ndb_select_count from 6.3.10 with libndbclient.so from 6.3.11 cause crash
Submitted: 16 Apr 11:45 Modified: 9 Jul 10:48
Reporter: Jonas Oreland
Status: Closed
Category:Server: Cluster Severity:S3 (Non-critical)
Version:6.3.11 OS:Any
Assigned to: Frazer Clement Target Version:

[16 Apr 11:45] Jonas Oreland
Description:
.

How to repeat:
.

Suggested fix:
.
[25 Apr 17:20] Frazer Clement
WL4086 modified the internals of scan and index scan so that they always use NdbRecord
mechanisms internally.
NdbRecord mechanisms support the NdbInterpretedCode Api rather than the old interpreted
program Api methods.
The NdbRecord checks in the old interpreted program Api methods were not being hit as they
were placed too late in the decision path.
This resulted in corrupt signals beings sent to the kernel.
[28 Apr 16:18] Bugs System
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/46124

ChangeSet@1.2588, 2008-04-28 15:17:28+01:00, frazer@forth.ndb.mysql.com +4 -0
  Bug #36124 
  Checks on operation type were not correct, resulting in corrupted signals being sent
when the 
  old interpreted code Api is used with scans.
[28 Apr 18:23] Bugs System
Pushed into 5.1.23-ndb-6.4.0
[28 Apr 18:25] Bugs System
Pushed into 5.1.24-ndb-6.3.13
[28 Apr 18:29] Bugs System
Pushed into 5.1.24-ndb-6.2.14
[9 Jul 10:48] Jon Stephens
Documented in the 5.1.23-ndb-6.2.14 and 5.1.24-ndb-6.3.13 changelogs as follows:

        Using NDB API applications from older MySQL Cluster versions with 
        libndbclient from newer ones caused the cluster to fail.