[
https://issues.apache.org/jira/browse/HBASE-2578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12871369#action_12871369
]
stack commented on HBASE-2578:
------------------------------
First, do we have examples of tests that are prone to the condition described
above in the description, and if so, what are they? Are they so many it merits
a system such as Dan has cast? Do we think that Dan's system will be filled
out further? If so, we should take a step back and try and work in some
general-purpose tooling (Since we're changing a fundamental,
currentSystemMillis() just in testing, it seems something like aspectj would
make sense in here).
If we think this a one-time only thing, then Dan's patch looks good to me.
> Add ability for tests to override server-side timestamp setting
> (currentTimeMillis)
> -----------------------------------------------------------------------------------
>
> Key: HBASE-2578
> URL: https://issues.apache.org/jira/browse/HBASE-2578
> Project: HBase
> Issue Type: Improvement
> Reporter: Jonathan Gray
> Assignee: Daniel Ploeg
> Attachments: HBASE-2578.txt
>
>
> Many of our tests use client APIs which do not set explicit stamps. This
> creates weird timing issues with tests running on different systems because
> sometimes a set of operations happens in the same millisecond and other times
> they do not.
> We should have a way for a test to specify it's own way of generating the
> timestamps (for example, could always increment by 1 ensuring forward
> progression in time).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
