Re: [DISCUSS] abandon branch protection rules

I feel that we should keep it but that we need to look into what's causing the frustration with the stresstest job.  That seems to be the thing causing the most grief.  People make a small change to some test, such as changing an import statement, and then find that it fails in stresstest. I

Re: [DISCUSS] abandon branch protection rules

@Jacob Barrett I have some ideas on this. Want to look at in on Thursday with me? On Sat, Dec 28, 2019 at 5:28 AM Jacob Barrett wrote: > Let’s find a way to get the ci, docs, and other directories not effected > by tests out of this testing hold. > > > On Dec 27, 2019, at 3:23 PM, Nabarun Nag

Re: [DISCUSS] abandon branch protection rules

@Jacob Barrett @Robert Houghton I have an interest in expediting docs-only PRs and would be interested in participating in a break-out discussion (or at least reviewing the conclusions). On Mon, Dec 30, 2019 at 9:15 AM Robert Houghton wrote: > @Jacob Barrett I have some ideas on this. Want to

Re: [DISCUSS] `status locator` command fail when locator's ssl is turned on

Long before we open-sourced as Geode, GemFire engineering planned to remove support for: gfsh>status locator --host=fooish --port=10334 The reasoning at the time was (1) this method for getting details about the locator was wide open (security hole), and (2) it's inconsistent with "status server"

Re: [DISCUSS] `status locator` command fail when locator's ssl is turned on

My vote is for #2 in Jinmei's proposal with the following correction: After removing --host/--port, you still have two ways to get "status locator": a) Use --dir option: gfsh>status locator --dir=/locator-dir b) Use connect command first: gfsh>connect locator gfsh>status locator --name=locato

Re: [DISCUSS] `status locator` command fail when locator's ssl is turned on

If we can remove these support, I would like this approach too. Current implementation has too many inconsistencies. On Mon, Dec 30, 2019 at 9:34 AM Kirk Lund wrote: > My vote is for #2 in Jinmei's proposal with the following correction: > > After removing --host/--port, you still have two ways

Re: [DISCUSS] abandon branch protection rules

Here's my take on stresstest. It's currently providing two purposes: 1) Prevents addition of new flaky tests Some new flakiness does slip through. I can write a new test that passes 50-100 times consistently and thus gets by stresstest, but then fails once or twice in CI or dunitrunner when I run

Re: [DISCUSS] `status locator` command fail when locator's ssl is turned on

Our first step is probably to deprecate it. We have some gfsh commands that have deprecated info in the help (search for deprecated in CliStrings.java) mostly for options we want to remove. On Mon, Dec 30, 2019 at 9:37 AM Jinmei Liao wrote: > If we can remove these support, I would like this app

Re: [DISCUSS] abandon branch protection rules

Regarding the StressTest job - how about we switch that have a new JVM for each test? That's how DistributedTest and IntegrationTest normally run. We let StressTest reuse the JVMs because it would be faster and find problems related to static state left behind, but I think in practice people are fi

Re: [DISCUSS] abandon branch protection rules

Having the stressTest reuse the JVMs is close to running the tests in my IDEA repeatedly for N times or running a package of tests together in my IDEA. There was a time that I couldn't run a group of tests together in my IDEA until I had to fix the problem for the stressTest. Keeping them running i

Passed: apache/geode-examples#396 (rel/v1.11.0 - 9e6f13e)

Build Update for apache/geode-examples - Build: #396 Status: Passed Duration: 23 mins and 56 secs Commit: 9e6f13e (rel/v1.11.0) Author: Mark Hanson Message: temporarily point to staging repo for CI purposes View the changeset: https://github.com/apache/geode-e

Re: [DISCUSS] abandon branch protection rules

We could keep it in one JVM but make it optional for merging the PR? On Mon, Dec 30, 2019 at 10:06 AM Jinmei Liao wrote: > Having the stressTest reuse the JVMs is close to running the tests in my > IDEA repeatedly for N times or running a package of tests together in my > IDEA. There was a time

Passed: apache/geode-native#2286 (rel/v1.11.0 - 3199cae)

Build Update for apache/geode-native - Build: #2286 Status: Passed Duration: 1 hr, 16 mins, and 58 secs Commit: 3199cae (rel/v1.11.0) Author: Jacob Barrett Message: GEODE-7426: Fixes segfault in log message. (#545) (cherry picked from commit 55da853760c200c53

Re: [DISCUSS] Proposal to require linear commit history on develop

Given that we adopted and still wish to continue

Re: [DISCUSS] Proposal to require linear commit history on develop

This change to disable all but squash-merge would be really easy to revert. How about we try it for a while and see? If people decide it is really limiting them, we can change it back. Let’s do it for 1 month and see how it goes. Does that sound reasonable? Thanks, Mark > On Dec 30, 2019, at 5