Re: [DISCUSS] Pre-commit tests before commits

2016-10-25 Thread Siddharth Seth
Added the following. https://cwiki.apache.org/confluence/display/Hive/HowToCommit#HowToCommit-PreCommitruns,andcommittingpatches On Tue, Oct 25, 2016 at 12:03 AM, Lefty Leverenz wrote: > > > > @Prasanth - separate page on the wiki for 'Adding / modifying tests'. > That can > > be discussed on th

Re: [DISCUSS] Pre-commit tests before commits

2016-10-25 Thread Peter Vary
Created a jira: HIVE-15051 Test framework integration with findbugs, rat checks etc. If there is someone who wants this task feel free to take it, otherwise I will do it myself. If you have any suggestions, ideas please comment. I will update my

Re: [DISCUSS] Pre-commit tests before commits

2016-10-25 Thread Lefty Leverenz
> > @Prasanth - separate page on the wiki for 'Adding / modifying tests'. That can > be discussed on the list, or modified directly. Here's what the wiki has so far: - Tips for Adding New Tests (written in 2011)

Re: [DISCUSS] Pre-commit tests before commits

2016-10-24 Thread Siddharth Seth
There seems to be general consensus to follow this approach for commits, and waiting for pre-commit tests. I'm going to add this to the Hive Wiki. @Alan, On using Yetus - I don't really know much about it, so cannot comment. However, Hive does use a parallel test framework which generates reports,

Re: [DISCUSS] Pre-commit tests before commits

2016-10-17 Thread Sergey Shelukhin
+1 On 16/10/13, 23:10, "Siddharth Seth" wrote: >There's been a lot of work to make the test runs faster, as well as more >reliable via HIVE-14547, HIVE-13503, and several other jiras. Test >runtimes >are around the 1 hour mark, and going down. There were a few green >pre-commit runs (after years

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Alan Gates
As far connecting in findbugs, etc. should we consider integrating with Apache Yetus for all this? Alan. > On Oct 14, 2016, at 12:27, Siddharth Seth wrote: > > Once we agree upon what the flow should be, I'm also in favor of reverting > patches which break things. Such commits cause problems f

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Peter Vary
About formatting changes we could do it with at least 2 different approach: Hadoop way - the precommit script which only analyzes the code, and raises a red flag if there is more error in the patched file than before the patch - no extra effort needed for it other than creating the infrastructure

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Chris Drome
+1. We have a build/test environment for branch-1.2 and found the following two reasons for some of the flakiness in tests: 1. Derby leaves its database files hanging around the target directory, so the next test framework (e.g. TestMinimrCliDriver, TestMiniTezCliDriver) that runs picks up the

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Eugene Koifman
Making global formatting changes will create a patch that changes almost every line and make git annotate useless which makes figuring out history of changes difficult. Perhaps disabling tests should require a separate +1 (ideally from initial author). On 10/14/16, 12:27 PM, "Siddharth Seth" wr

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Siddharth Seth
Once we agree upon what the flow should be, I'm also in favor of reverting patches which break things. Such commits cause problems for all subsequent patches - where multiple people end up debugging the test. In terms of disabling tests - we may want to consider how often the test fails, before dis

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Peter Vary
+1 from me too. I think it lowers the barrier for the contributors, if there are clean and easy to follow rules for adding a patch. I had very good experience with the Hadoop commit flow where only “all green” results are accepted. If we are thinking about changes in the commit flow, it would b

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Zsombor Klara
+1 from my side as well. I also like the suggestion to revert a patch causing new test failures instead of expecting the owner to fix the issues in follow up jiras. And I would also have a borderline "radical" suggestion.. if a flaky test is failing and the committer isn't sure at a glance that th

Re: [DISCUSS] Pre-commit tests before commits

2016-10-14 Thread Prasanth Jayachandran
+1 on the proposal. Adding more to it. A lot of time has been spent on improving the test runtime and bringing down the flaky tests. Following jiras should give an overview of the effort involved https://issues.apache.org/jira/browse/HIVE-14547 https://issues.apache.org/jira/browse/HIVE-13503 Co

[DISCUSS] Pre-commit tests before commits

2016-10-13 Thread Siddharth Seth
There's been a lot of work to make the test runs faster, as well as more reliable via HIVE-14547, HIVE-13503, and several other jiras. Test runtimes are around the 1 hour mark, and going down. There were a few green pre-commit runs (after years?). At the same time, there's still some flaky tests.