-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/47429/#review133765
-----------------------------------------------------------


Ship it!




I agree with the original reasoning for naming method isSmallEnough. The test 
is for whether or not the combined size of adjacent free fragments is too big 
or too small to still be a valid fragment size; so if it 'isSmallEnough' then 
combine them.

- Ken Howe


On May 17, 2016, 11:13 p.m., Darrel Schneider wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/47429/
> -----------------------------------------------------------
> 
> (Updated May 17, 2016, 11:13 p.m.)
> 
> 
> Review request for geode, Eric Shu, Scott Jewell, Ken Howe, and Sai 
> Boorlagadda.
> 
> 
> Bugs: GEODE-1292
>     https://issues.apache.org/jira/browse/GEODE-1292
> 
> 
> Repository: geode
> 
> 
> Description
> -------
> 
> change defragment to not create fragments > 2G
> 
> 
> Diffs
> -----
> 
>   
> geode-core/src/main/java/com/gemstone/gemfire/internal/offheap/FreeListManager.java
>  c5e17e860e8e3300aac490621ae5d5b50f3a0e38 
>   
> geode-core/src/main/java/com/gemstone/gemfire/internal/offheap/OffHeapStoredObjectAddressStack.java
>  b69d3a64cb09da953d196e112a6c635f80df0abb 
>   
> geode-core/src/main/java/com/gemstone/gemfire/internal/offheap/SlabImpl.java 
> 1c88bde9379c5408e15c4764da10db9ddb08a352 
>   
> geode-core/src/test/java/com/gemstone/gemfire/internal/offheap/FreeListManagerTest.java
>  4cf1df00bb63154da03184c77f7656dbd892bbbb 
> 
> Diff: https://reviews.apache.org/r/47429/diff/
> 
> 
> Testing
> -------
> 
> precheckin
> 
> I looked over the code to see if I could create a unit test that does not 
> need to allocate 2G chunks of off-heap memory and it was not possible without 
> major refactoring. The problem is that we modify size fields stored at an 
> address during defragmentation and currently that requires that the address 
> be a valid off-heap memory address.
> 
> 
> Thanks,
> 
> Darrel Schneider
> 
>

Reply via email to