Did you read the issue at all? You have to run the test thousands of times with a particular random seed to have a chance of reproducing.
On Sun, Dec 7, 2014 at 5:19 AM, Michael Wechner <michael.wech...@wyona.com> wrote: > Hi Robert > > Thanks very much for your feedback. I guess you mean the following change: > > git diff > 1f69438a17a0ad33bb9c209113b0fe83fb68f860..5bb9e0aebe18a67e5e0e4698e0553d6824327495 > lucene/core/src/java/org/apache/lucene/index/IndexWriter.java > > - assert lastCommitChangeCount <= changeCount: > "lastCommitChangeCount=" + lastCommitChangeCount + " changeCount=" + > changeCount; > + if (lastCommitChangeCount > changeCount) { > + throw new IllegalStateException("lastCommitChangeCount=" + > lastCommitChangeCount + ",changeCount=" + changeCount); > + } > > right? > > I really did not mean to point fingers, but I just try to think of a > "testing environment" which assists developers as good and as efficient > as possible to find things which might have been forgotten or maybe > break something. > I think the above is a good example how difficult it can be to do this, > but still I think it is possible and would help us all. > > Hence I would like to ask again, what is the general rule for updating > CHANGES.txt > > About missing tests, I guess you experienced a buggy G1GC collector or > other JVM bugs and hence you > made this change, right? If so, I guess you will run the new code with > the existing buggy G1GC collector > and if an IllegalStateException is being thrown, then you will know that > something is wrong, right? > Or if I misunderstand, can you explain a bit more? > > Thanks > > Michael > > > Am 07.12.14 um 10:29 schrieb Robert Muir: >> Look at the actual change in question, >> >> It is not "missing" tests and cannot be tested. >> >> It is a guard against jvm bugs like buggy G1GC collector. >> >> On Sat, Dec 6, 2014 at 11:25 PM, Michael Wechner >> <michael.wech...@wyona.com> wrote: >>> Hi >>> >>> Yesterday "morning" I have noticed the following changes >>> >>> - lucene/core/src/java/org/apache/lucene/index/ConcurrentMergeScheduler.java >>> - lucene/core/src/java/org/apache/lucene/index/MergePolicy.java >>> - >>> lucene/core/src/test/org/apache/lucene/index/TestConcurrentMergeScheduler.java >>> >>> and with these modifications also CHANGES.txt was updated. >>> >>> Yesterday "afternoon" I have noticed the following change >>> >>> - lucene/core/src/java/org/apache/lucene/index/IndexWriter.java >>> >>> but this time CHANGES.txt was not updated (also >>> lucene/core/src/test/org/apache/lucene/index/TestIndexWriter.java was not >>> updated). >>> >>> As written in a previous email I am trying to test changes incrementally, >>> hence I was wondering whether it would be helpful, that such a system for >>> incremental testing would also point out things like maybing missing >>> updates, like for example for CHANGES.txt or associated tests. >>> >>> But in order to do so, the system should implement some kind of rule, hence >>> I was wondering what the rule is about CHANGES.txt? >>> >>> I know it sounds a bit like parents saying "hey, have you brushed your >>> teeth?!" and it can be rather annoying, but it can also be very useful for >>> the future having good teeth :-) >>> >>> Thanks >>> >>> Michael >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org >> For additional commands, e-mail: dev-h...@lucene.apache.org >> > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > For additional commands, e-mail: dev-h...@lucene.apache.org > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org