[ 
https://issues.apache.org/jira/browse/SPARK-61471?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Hyukjin Kwon updated SPARK-61471:
---------------------------------
    Description: Follow-up of SPARK-61400. A DataFrame of 300 withColumn levels 
over an expression of 700 additions crashed the process with SIGBUS when its 
schema was read on a thread of 512 KiB, although each depth is within its own 
bound: with the plan compression of a real server, the encoded size of the plan 
is measured on the thread of the call. SPARK-61469 (#522) fixes the crash by 
driving the future of a blocking call from the heap. This adds a regression 
test of the combined shape.  (was: Follow-up of SPARK-61400. A DataFrame of 300 
withColumn levels over an expression of 700 additions crashes the process with 
SIGBUS when its schema is read on a thread of 512 KiB, although each depth is 
within its own bound. The clone of the request runs on a grown stack, but the 
rest of the analysis request does not.)
     Issue Type: Test  (was: Bug)
        Summary: Test that a deep plan over a deep expression is analyzed from 
a thread with a small stack  (was: Keep the whole request of a deep analysis on 
a grown stack, so that a deep plan with a deep expression does not crash a 
thread with a small stack)

> Test that a deep plan over a deep expression is analyzed from a thread with a 
> small stack
> -----------------------------------------------------------------------------------------
>
>                 Key: SPARK-61471
>                 URL: https://issues.apache.org/jira/browse/SPARK-61471
>             Project: Spark
>          Issue Type: Test
>          Components: Connect
>    Affects Versions: connect-rust-4.3.0
>            Reporter: Hyukjin Kwon
>            Priority: Major
>             Fix For: connect-rust-4.3.0
>
>
> Follow-up of SPARK-61400. A DataFrame of 300 withColumn levels over an 
> expression of 700 additions crashed the process with SIGBUS when its schema 
> was read on a thread of 512 KiB, although each depth is within its own bound: 
> with the plan compression of a real server, the encoded size of the plan is 
> measured on the thread of the call. SPARK-61469 (#522) fixes the crash by 
> driving the future of a blocking call from the heap. This adds a regression 
> test of the combined shape.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to