Dan Hecht has posted comments on this change.

Change subject: IMPALA-5173: crash with hash join feeding directly into nlj
......................................................................


Patch Set 2:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/6568/2/be/src/exec/nested-loop-join-builder.cc
File be/src/exec/nested-loop-join-builder.cc:

Line 55:     // data in that cases. TODO: remove workaround when IMPALA-4179 is 
fixed
this is so confusing.  here you say we can't acquire ownership of these, yet 
TransferResourceOwnership() moves these blocks around. so are they transferred 
or not?

Remind me why blocks are attached to the row-batch if they can't get 
transferred? is it basically to defer/control the timing of the deletion of the 
block? and why is this different than the MarkNeedsDeepCopy() case? 

and why isn't this a problem for other cases that check the needs_deep_copy() 
flag? oh, is this the only place we have to check for that?


-- 
To view, visit http://gerrit.cloudera.org:8080/6568
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I0c04952e591d17e5ff7e994884be4c4c899ae192
Gerrit-PatchSet: 2
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Tim Armstrong <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-HasComments: Yes

Reply via email to