[
https://issues.apache.org/jira/browse/TINKERPOP3-885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14964760#comment-14964760
]
Dylan Millikin commented on TINKERPOP3-885:
-------------------------------------------
Yes this fixes the test. I've been holding out because I wanted to run the full
test suit before making a PR. Just to make sure we aren't breaking something.
I also realized there was a bit of a misunderstanding in that I never actually
manually changed the transaction management. So the bug I had and reported the
other day wasn't due to that. But I'll take a closer look at the problem I was
experiencing and we can handle that separately.
> Change Transaction.onReadWrite() to be a ThreadLocal setting
> ------------------------------------------------------------
>
> Key: TINKERPOP3-885
> URL: https://issues.apache.org/jira/browse/TINKERPOP3-885
> Project: TinkerPop 3
> Issue Type: Improvement
> Components: server, structure
> Affects Versions: 3.0.1-incubating
> Reporter: Dylan Millikin
> Assignee: Dylan Millikin
> Labels: breaking
> Fix For: 3.1.0-incubating
>
>
> The issue is as follows:
> Because Transaction consumers are global for a graph. A
> {{SessionOpProcessor}} request will change the Transaction read write
> behavior to {{MANUAL}} across all concurrent/future requests.
> This will make other requests fail (ones expecting {{AUTO}}).
> This has been discussed here :
> http://mail-archives.apache.org/mod_mbox/incubator-tinkerpop-dev/201510.mbox/browser
> The solution would be to make Transaction consumers ThreadLocal to keep the
> state local to the requests.
> This is a breaking change.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
