[
https://issues.apache.org/jira/browse/LUCENE-3195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13048379#comment-13048379
]
Robert Muir commented on LUCENE-3195:
-------------------------------------
{quote}
Also, and this can be added in future iterations, I think it'd be good if the
exception includes a list of all test classes + count of open files, to make it
easier to spot offending classes.
{quote}
Well, the test spits this out when it fails already? it says "tests run in this
jvm: [TestX, TestY, TestZ]..."
{quote}
Hmmm ... now that I think of it, if our tests are run in different JVMs, we
will fail to faithfully track the number of open file handles (I mean the true
number)? In that case, we might want to lower the value to whatever we think a
single JVM should allocate ... which gets tricky.
{quote}
Right, its not going to be totally accurate anyway, I guess I think the goal is
not to emulate an operating system but instead to catch a runaway test on all
platforms.
For example, as a start MDW could just check its own hashmap, as 99% of tests
only make one directory, and a JVM only runs one test at a time.
But it would be better for MDW to have a static atomic counter, for better
detection of the crazy tests (those are probably the ones opening too many
files sometimes anyway).
> MockDirectoryWrapper should throw 'too many open files' if too many are open
> ----------------------------------------------------------------------------
>
> Key: LUCENE-3195
> URL: https://issues.apache.org/jira/browse/LUCENE-3195
> Project: Lucene - Java
> Issue Type: Test
> Reporter: Robert Muir
>
> regardless of operating system.
> We should take a good default value, like the lowest one we consider
> 'reasonable' macosX? and it should throw exception on openInput if openFiles
> hashmap is too large.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
