Bug #36389 Get Ndbif: Error Ndb::handleReceivedSignal when running debug build
Submitted: 29 Apr 2008 2:21 Modified: 13 Aug 2008 20:55
Reporter: Monty Taylor Email Updates:
Status: Won't fix Impact on me:
None 
Category:MySQL Cluster: NDB API Severity:S3 (Non-critical)
Version:6.2 or 6.3 OS:Any
Assigned to: Gustaf Thorslund CPU Architecture:Any

[29 Apr 2008 2:21] Monty Taylor 
Description:
When running a debug version of 6.2 or 6.3, I get 

Ndbif: Error Ndb::handleReceivedSignal (tFirstDataPtr=0x64e190, GSN=10, theImpl->theWaiter.m_state=0) sender = (Block: 245 Node: 1)

If I execute a transaction that contains a duplicate key error. 

The problem seems to be (guess, looking at trace log) a TCKEYCONF signal sent after TCROLLBACKREP. 

How to repeat:
Compile the attached test program. Make sure you are compiling against a recent 6.2 or 6.3 built --with-debug=full. 

Suggested fix:
I have absolutely no clue. Either don't send the signal, or handle the signal properly?
[29 Apr 2008 2:24] Monty Taylor 
Test program showing the problem.

Attachment: testabort.cpp (text/x-c++src), 2.87 KiB.
[29 Apr 2008 2:25] Monty Taylor 
output from running the program with API_TRACE

Attachment: testabort.out (application/octet-stream, text), 10.08 KiB.
[2 May 2008 13:10] Hartmut Holzgraefe 
All i'm getting from this is:

Error in testabort.cpp, line: 106, code: 630, msg: Tuple already existed when attempting to insert.

not sure what is wrong there, the test table gets created but still empty after the program terminated ...
[5 May 2008 9:39] Jonas Oreland 
hi,

i analyzed this before and concluded that it was benign,
but a lot of work to fix.

i.e "to be fixed later" where later = 'when someone feels like fighting with tc'

---

i you can find any incorrect results, then it's of course a different story
[13 Mar 2014 13:36] Omer Barnir 
This bug is not scheduled to be fixed at this time.