RFR: 8224974: Implement JEP 352

Dmitry Chuyko dmitry.chuyko at bell-sw.com
Wed Aug 7 11:02:47 UTC 2019


New code is buildable and new MapFail test passes on different platforms 
except it fails on linux i386:

java.lang.Exception: unexpected message for IOExceptionInvalid argument
     at MapFail.main(MapFail.java:60)
     at java.base/java.lang.reflect.Method.invoke(Method.java:565)
     at java.base/java.lang.Thread.run(Thread.java:830)

First there is a problem with the test,
and a minor test issue is it would be good to add ": " before actual 
unexpected message.


On 8/7/19 12:31 PM, Dmitry Chuyko wrote:
> On 8/6/19 6:58 PM, Andrew Dinn wrote:
>> ......................
>> No this behaviour is not currently tested. However, the only client at
>> present will never exercise that path so it is not critical to test it
>> now. I'd be happy to address testing of this behaviour as part of a
>> follow-up JIRA issue if you would be so good as to raise it. I say that
>> because I would very much like to get this functionality into a release
>> to simplify more extensive testing by Red Hat's middleware teams.
> It sounds reasonable, I'll create a tiny RFE after you push the JEP.
>>> New MapFail test succeeds if proper IOException or any
>>> UnsupportedOperationException was caught but it aren't those situations
>>> actually 2 different ones that require distinct checks? If you say that
>>> is the situation when results depend on Linux version, it makes 
>>> sense at
>>> least to put a comment in the test because it's really not trivial.
>> The documentation of the API under test makes it clear that both errors
>> can occur and under what circumstances. However, a note in the test will
>> certainly do no harm. I will insert one before checking in the patch.
>>> Can PmemTest check IOException with message "map with mode MAP_SYNC
>>> unsupported" as a part of expected behavior, not just showing a test
>>> failure?
>> I don't see any need for this now that MapFail has been provided. Wit
>> that alterative in place for checking map failures on non-DAX file
>> syetems PmemTest is now primarily intended to test behaviour with a DAX
>> file system which it expects to have been set up in advance as described
>> in the main comment. So, the scenario you describe is not really an
>> intended usage and I woudl argue that a failure is the right way to
>> signal that.
> OK, finally got it, thank you.
> -Dmitry
>> regards,
>> Andrew Dinn
>> -----------
>> Senior Principal Software Engineer
>> Red Hat UK Ltd
>> Registered in England and Wales under Company Registration No. 03798903
>> Directors: Michael Cunningham, Michael ("Mike") O'Neill, Eric Shander

More information about the core-libs-dev mailing list