advice & review requested for 6896297, race condition in rmid causing JCK failure

Alan Bateman Alan.Bateman at
Wed Apr 13 14:54:58 UTC 2011

Stuart Marks wrote:
> :
> The key issue seems to be whether we want to preserve compatibility of 
> the serialized form. If we care about compatibility of the serialized 
> form, #3 is probably the safest. If somebody comes along and tells us 
> that we don't have to worry about serial compatibility of this stuff 
> at all, then #1 is probably the best since it simplifies the code a lot.
I think (and Peter can say for sure) that the compatibility issue is 
really only for the case that rmid needs to recover from an existing log 
file. Without more insight then I tend to agree that #3 is the best for now.


More information about the core-libs-dev mailing list