RFR(S) : 8181053 : port basicvmtest to jtreg
erik.helin at oracle.com
Tue Jun 27 06:45:34 UTC 2017
looking at this one extra time, I realized that the following change to
-# clienttest (make sure various basic java client options work)
-hotspot_clienttest clienttest: sanitytest
- $(RM) $(PRODUCT_HOME)/jre/lib/*/client/classes.jsa
- $(RM) $(PRODUCT_HOME)/jre/bin/client/classes.jsa
- $(PRODUCT_HOME)/bin/java $(JAVA_OPTIONS) -Xshare:dump
-PHONY_LIST += hotspot_clienttest clienttest
actually removes one additional test that isn't covered by your newly
added file hotspot/test/sanity/BasicVMTest.java: sanity testing
Do we want to add a small test in hotspot/test/sanity for -Xshare:dump?
Or is this functionality tested elsewhere? A related question: if
multiple tests were running concurrently (testing the same JDK), won't
there be a race condition with the above test? For example if two JTReg
tests are running (and -conc > 1) and both JTReg tests tries to remove
classes.jsa and then regenerate them, seems like there could be a race?
What do you think? Just scrap the -Xshare:dump sanity test or add a
On 06/14/2017 09:41 AM, Erik Helin wrote:
> On 06/14/2017 01:09 AM, Igor Ignatyev wrote:
>>> 121 lines changed: 54 ins; 67 del; 0 mod;
>> Hi all,
>> could you please review this small patch which introduces jtreg
>> version of basicvmtest test?
>> make version of basicvmtest also included sanity testing for CDS on
>> client JVM, but this testing modified the product binaries, so it
>> might interfere with results of other tests and is not very reliable.
>> I have consulted w/ Misha about this, and he assured me that there are
>> other better CDS tests which check the same functionality, so we
>> should not lose test coverage there.
> Looks good, Reviewed.
> Thank you for this patch Igor! I've been meaning to fix this for a long
> time but never got around to it...
>> jbs: https://bugs.openjdk.java.net/browse/JDK-8181053
>> testing: jprt, new added test
>> -- Igor
More information about the hotspot-dev