I don't think holding up the release process infinitely till we stabilize all the tests is an option. On the other hand getting an RC to build is pretty difficult ( I am facing the same problem with 6.6.1 ) and I am sure people will run into this while voting for the release?
We could identify the top 2/3 tests which fail regularly while building the RC and either disable them or see if someone volunteers to fix them ? On Tue, Aug 29, 2017 at 2:53 AM, Ishan Chattopadhyaya < [email protected]> wrote: > > Those flaky Solr tests are annoying since people will also run into > failures when > > checking the RC? Should we disable these tests on the 7.0 branch so that > building > > and verifying this RC isn't annoying to everybody working on this > release? > > +1. If it is hampering the release process, I think we should either not > release without fixing them, or disable them for release (building, > verifying). > > On Mon, Aug 28, 2017 at 11:47 PM, Anshum Gupta <[email protected]> wrote: > >> Though those failing tests are annoying, I would not recommend ignoring >> those tests. We can manually ignore those test failures when we are testing >> stuff out though. >> >> -Anshum >> >> >> >> On Aug 28, 2017, at 11:10 AM, Adrien Grand <[email protected]> wrote: >> >> Those flaky Solr tests are annoying since people will also run into >> failures when checking the RC? Should we disable these tests on the 7.0 >> branch so that building and verifying this RC isn't annoying to everybody >> working on this release? >> >> Le lun. 28 août 2017 à 19:23, Anshum Gupta <[email protected]> a écrit : >> >>> Thanks Adrien! It worked with a fresh clone, at least ant check-licenses >>> worked, so I’m assuming the RC creation would work too. >>> I’m running that, and it might take a couple of hours for me to create >>> one, as a few SolrCloud tests are still a little flakey and they fail >>> occasionally. >>> >>> -Anshum >>> >>> >>> >>> On Aug 28, 2017, at 10:13 AM, Anshum Gupta <[email protected]> wrote: >>> >>> Adrien, >>> >>> Yes, ant check-licenses fails with the same error, and so does ant >>> validate (from the root dir). This is after running ant clean -f. >>> >>> BUILD FAILED >>> /Users/anshum/workspace/lucene-solr/build.xml:117: The following error >>> occurred while executing this line: >>> /Users/anshum/workspace/lucene-solr/lucene/build.xml:90: The following >>> error occurred while executing this line: >>> /Users/anshum/workspace/lucene-solr/lucene/tools/custom-tasks.xml:62: >>> JAR resource does not exist: analysis/icu/lib/icu4j-56.1.jar >>> >>> I didn’t realize that the dependency was upgraded, and what confuses me >>> is that the file actually exists. >>> >>> anshum$ ls analysis/icu/lib/icu4j-5 >>> icu4j-56.1.jar icu4j-59.1.jar >>> >>> It seems like it’s something that git clean, ant clean clean-jars etc. >>> didn’t fix. This is really surprising but I’ll try and checking out again >>> and creating and RC (after checking for the dependencies). >>> I think ant should be responsible for cleaning this up, and not git so >>> there’s something off there. >>> >>> -Anshum >>> >>> >>> >>> On Aug 28, 2017, at 8:51 AM, Adrien Grand <[email protected]> wrote: >>> >>> You mentioned you tried to run the script multiple times. Have you run >>> git clean at some point? Maybe this is due to a stale working copy? >>> >>> Le lun. 28 août 2017 à 08:53, Adrien Grand <[email protected]> a écrit : >>> >>>> Hi Anshum, >>>> >>>> Does running ant check-licenses from the Lucene directory fail as well? >>>> The error message that you are getting looks weird to me since Lucene 7.0 >>>> depends on ICU 59.1, not 56.1 since https://issues.apache.org/jira >>>> /browse/LUCENE-7540. >>>> >>>> Le ven. 25 août 2017 à 23:42, Anshum Gupta <[email protected]> a >>>> écrit : >>>> >>>>> A quick question, in case someone has an idea around what’s going on. >>>>> When I run the following command: >>>>> >>>>> python3 -u dev-tools/scripts/buildAndPushRelease.py --push-local >>>>> /Users/anshum/solr/release/7.0.0/rc0 --rc-num 1 --sign <my-key> >>>>> >>>>> I end up with the following error: >>>>> >>>>> BUILD FAILED >>>>> /Users/anshum/workspace/lucene-solr/build.xml:117: The following >>>>> error occurred while executing this line: >>>>> /Users/anshum/workspace/lucene-solr/lucene/build.xml:90: The >>>>> following error occurred while executing this line: >>>>> /Users/anshum/workspace/lucene-solr/lucene/tools/custom-tasks.xml:62: >>>>> JAR resource does not exist: analysis/icu/lib/icu4j-56.1.jar >>>>> >>>>> Any idea as to what’s going on? This generally fails after the tests >>>>> have run, and the script has processed for about 45 minutes and it’s >>>>> consistent i.e. all the times when the tests pass, the process fails with >>>>> this warning. >>>>> >>>>> I can also confirm that this file exists at >>>>> lucene/analysis/icy/lib/icu4j-56.1.jar . >>>>> >>>>> Has anyone else seen this when working on the release? >>>>> >>>>> -Anshum >>>>> >>>>> >>>>> >>>>> On Aug 23, 2017, at 4:21 AM, Andrzej Białecki < >>>>> [email protected]> wrote: >>>>> >>>>> >>>>> On 23 Aug 2017, at 13:06, Uwe Schindler <[email protected]> wrote: >>>>> >>>>> Keep in mind that there is also branch_7_0. >>>>> >>>>> >>>>> Right, but the changes related to these issues were committed to >>>>> master before branch_7_0 was created, and these specific issues are only >>>>> about back-porting to 6x. >>>>> >>>>> >>>>> Uww >>>>> >>>>> Am 23. August 2017 12:26:42 MESZ schrieb "Andrzej Białecki" < >>>>> [email protected]>: >>>>>> >>>>>> >>>>>> On 23 Aug 2017, at 08:15, Anshum Gupta <[email protected]> wrote: >>>>>> >>>>>> I also found more issues when comparing 7x, with 6x this time. I’ll >>>>>> take a look at wether it’s just the CHANGES entries or have these >>>>>> actually >>>>>> missed the branch. I assume it’s just the CHANGES, but want to be sure. >>>>>> If >>>>>> the committers involved can pitch in, I’d appreciate, else I’ll work on >>>>>> this for a bit right now and continue with this tomorrow morning. >>>>>> >>>>>> - SOLR-10477 (Ab) >>>>>> >>>>>> >>>>>> This is a partial back-port of relevant improvements from master to >>>>>> 6x, so there are no strictly corresponding commits on 7x/master. >>>>>> >>>>>> - SOLR-10631: Metric reporters leak on 6x. (Ab) >>>>>> >>>>>> >>>>>> This one has been fixed as part of other related issues in branches >>>>>> 7.x / master, so it only required a specific fix for 6x. >>>>>> >>>>>> - SOLR-10000 (Ab) >>>>>> >>>>>> >>>>>> This has been committed first to 7x, then to 6x and it’s present in >>>>>> branch_6_6. >>>>>> >>>>>> >>>>>> --- >>>>>> Best regards, >>>>>> >>>>>> Andrzej Bialecki >>>>>> >>>>>> >>>>> -- >>>>> Uwe Schindler >>>>> Achterdiek 19, 28357 Bremen >>>>> https://www.thetaphi.de >>>>> >>>>> >>>>> >>>>> >>> >>> >> >
