[Date Prev][Date Next] [Chronological] [Thread] [Top]

(ITS#7144) reproducible segfault on slave/consumer during sync-replication

Full_Name: Sven Hartge
Version: 2.4.28
OS: Debian Squeeze
Submission from: (NULL) (2a01:198:329:1::14)

I experience an easy to reproduce and consistent segfault when I setup a slapd  
syncrepl consumer on Squeeze.                                                   
This segfault always happens during the replication of the same object, but the 
slapd version in Lenny has no problems with same DIT.                           
My oldstable replicas currently run a self-backported version of                
2.4.21-1~dvz50+1, but both 2.4.17-2.1~bpo50+1 and 2.4.11-1+lenny2.1 are also    
2.4.23-7.2 shows the segfault as all versions up to 2.4.28-1.1 do. Even if I    
disable all indices and strip my configuration to the bare minimum needed (i.e. 
self-defined objectclasses and attributes) I get the segfault.                  

I also did da -O0 rebuild of the Debian-2.4.28 and this also segfaults in the
same way as 2.4.23 did. I have yet to test the latest RelEng24 from GIT, but I
wanted to submit this issue so someone may have a look at the backtrace in
parallel to my further tests.
I am aware the problem my lie in my self-defined objectclasses and attributes,  
but then slapd should throw an error and exit instead of replicating 1/3 of the 
DIT and then segfault.                                                          
I get a very good backtrace, but since this backtrace contains internal         
information I am a bit hesitant to attach this backtrace to a public bug-report.
Is there a way to privatly submit this data (backtrace and additional schemas)  
so you can have look at the problem?