[ 
https://issues.apache.org/jira/browse/DISPATCH-121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14348751#comment-14348751
 ] 

Alan Conway edited comment on DISPATCH-121 at 3/5/15 1:54 PM:
--------------------------------------------------------------

disable_pools.diff: patch to allow disabling memory pooling for testing with 
valgrind. Use
    cmake -DUSE_VALGRIND=1 -DUSE_MEMORY_POOL=0

Not yet committed pending 0.4 release.


was (Author: aconway):
disable_pools.diff: patch to allow disabling memory pooling for testing with 
valgrind. Use
    cmake -DUSE_VALGRIND=1 -DUSE_MEMORY_POOL=0


> Alllow pool allocation to be switched for plain allocation at build time.
> -------------------------------------------------------------------------
>
>                 Key: DISPATCH-121
>                 URL: https://issues.apache.org/jira/browse/DISPATCH-121
>             Project: Qpid Dispatch
>          Issue Type: Improvement
>          Components: Container
>    Affects Versions: 0.3
>            Reporter: Alan Conway
>            Assignee: Alan Conway
>             Fix For: 0.5
>
>         Attachments: disable_pools.diff
>
>
> Memory checking tools like valgrind can't give much help when we use a custom 
> allocator.
> Make it possible to build dispatch  using malloc/free instead so it can be 
> checked with valgrind.
> This is not the ideal solution. It tests the code that uses the allocator but 
> does not test the allocator itself. 
> A better solution is to instrument the allocator for valgrind use, see:
> http://www.valgrind.org/docs/manual/mc-manual.html#mc-manual.mempools
> However  this is more complex and doesn't help with other memory debugging 
> tools. Disabling pools at build time is a useful first step. It also lets us 
> test performance of the pool allocator against plain malloc. A full valgrind 
> integration with pool allocation is a possible future exercise.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

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

Reply via email to