[ 
https://issues.apache.org/jira/browse/CASSANDRA-14775?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16815304#comment-16815304
 ] 

Alex Deparvu commented on CASSANDRA-14775:
------------------------------------------

I added more to the initial tests (and in the end had to pull in the patch from 
CASSANDRA-14773, code and tests though the tests were not technically needed). 
[0]
I separated out the failing tests so they can be looked at and also added some 
awkward fixes for the overflow just so I get the overall qt test passing. I did 
mark the code path that I saw breaking with TODOs so if anyone wants to take a 
look it's easy to spot. Interestingly enough not all possible overflows have 
been hit (not sure why, maybe the generator is not extensive enough).

I think there are also some trivial improvements that can be had with very low 
probability that something will break (like the {{DataHolder#size}} method or 
breaking early of the {{Spool#forEach}} if there are no items to be processed).

looking forward to the feedback!

[0] 
https://github.com/apache/cassandra/compare/trunk...stillalex:CASSANDRA-14775

> StreamingTombstoneHistogramBuilder overflows if > 2B in a single 
> bucket/stabile
> -------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-14775
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-14775
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Local/Compaction
>            Reporter: Benedict
>            Priority: Normal
>             Fix For: 4.0
>
>
> This may be unlikely, but is certainly not impossible.  In this event, the 
> count for the bucket will be reset to zero, and the time distorted to 1s in 
> the future.  If MAX_DELETION_TIME were encountered through overflow, this 
> might result in a bucket with NO_DELETION_TIME.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org

Reply via email to