RFR (S): 8024634 - gc/startup_warnings tests can fail due to unrelated warnings
per.liden at oracle.com
Mon Sep 23 12:54:59 UTC 2013
On 2013-09-23 14:23, Jesper Wilhelmsson wrote:
> The change look OK.
> Would the tests be more "stable" if they were checking for more than
> one word in the warning?
Problem is that the test wants to check that "nothing" is printed. Where
"nothing" here is a bit undefined, but at least it shouldn't say
anything about this or that GC option being deprecated. Checking for
more than one word doesn't help much because we actually don't know what
we are looking for. The whole test is a bit questionable, but with this
change it at least doesn't fail because of random unrelated warnings.
> Are there any tests that checks that we do get the deprecated warning
> for deprecated flag combinations? If so, they should probably be
> changed in the same way so that they don't pass because some other non
> related warning is issued.
Yes, there are such tests, but they are stable because unlike the other
tests these are checking that a specific warning _is_ printed, for example:
DefaultMaxRAMFraction is deprecated and will likely be removed in a
future release. Use MaxRAMFraction instead.");
TestDefNewCMS.java: output.shouldContain("warning: Using the DefNew
young collector with the CMS collector is deprecated and will likely be
removed in a future release");
TestCMSIncrementalMode.java: output.shouldContain("warning: Using
incremental CMS is deprecated and will likely be removed in a future
TestParNewSerialOld.java: output.shouldContain("warning: Using the
ParNew young collector with the Serial old collector is deprecated and
will likely be removed in a future release");
> Per Liden skrev 23/9/13 1:15 PM:
>> Could I please have a couple of reviews on this small test fix.
>> Summary: Some tests check that no warnings are printed when a
>> specific GC is
>> used. These test are a bit too generic and can catch and fail on all
>> sorts of
>> non-GC related warnings. What we actually want to test is that we
>> don't print
>> any message about these GCs being deprecated (as is the case for some
>> other GC
>> combinations). This isn't exactly bullet proof either since some non-GC
>> component could print something containing "deprecated", but if that
>> starts to
>> become a problem we'll have to reconsider the usefulness of these tests.
>> Testing: Have run the jtreg tests for GC
More information about the hotspot-gc-dev