[krbdev.mit.edu #7772] RE: krb5-1.11: Bug fix: iprop

[Richard Basch via RT]

Actually, it doesn't impact the current 1.11 because if this condition
occurs, the ulog last_sno will be set to 0, forcing a full resync on the
next attempt. 

 

From: Richard Basch [mailto:basch at alum.mit.edu] 
Sent: Saturday, November 02, 2013 10:04 PM
To: 'krb5-bugs at mit.edu'
Subject: krb5-1.11: Bug fix: iprop

 

I believe this patch applies to all 1.11.x releases, but certainly
manifested itself quite spectacularly with my iprop fixes to allow tree
replication.

 

Anyway, it is possible for a principal deletion event to be applied via
ulog_replay but not fully applied to the ulog (probably because of a lock
failure). Subsequently, when kpropd attempts to re-apply the missing
updates, it may re-attempt a deletion event and fail (add/modify events
work, but deletions can fail). The following patch fixes this.

 

https://github.com/rbasch/krb5/commit/cdd7b8c6bebbab29f1dedab40982b9d1672031
1e