On Tue, Aug 7, 2012 at 10:32 AM, Pavel Krivanek <pavel.krivanek@gmail.com>wrote:

We are starting to have a lot of failing tests again

CompressionTests.Archive.ZipArchiveTest.testCreateWithRelativeNames CompressionTests.Archive.ZipArchiveTest.testZip FileSystem.Tests.Disk.DiskFileSystemTest.testEntryAt FileSystem.Tests.Memory.MemoryFileSystemTest.testEntryAt KernelTests.Chronology.StopwatchTest.testSingleTiming KernelTests.Chronology.StopwatchTest.testStartStop KernelTests.Chronology.TimeStampTest.testComparing KernelTests.Chronology.TimeStampTest.testSorting Zinc.Tests.ZnLogSupportTests.testLogEvent RPackage.Tests.RPackageWithDoTest.testInvariant Tests.Release.ReleaseTest.testMethodsWithUnboundGlobals Tests.Release.ReleaseTest.testUndeclared

I think that we have to upgrade the monkey checking to stop such regressions...

Shouldn't Jenkins help us here?

-- Pavel

On Mon, Aug 6, 2012 at 4:27 PM, Esteban Lorenzano <estebanlm@gmail.com> wrote:

...

20245 -----

Issue 6510: update zinc http://code.google.com/p/pharo/issues/detail?id=6510

Issue 6514: Add allDefinedExtensionMethods to RPackage http://code.google.com/p/pharo/issues/detail?id=6514

Issue 6521: RPackageOrganizer>>initialize does not initialize from scratch http://code.google.com/p/pharo/issues/detail?id=6521

Issue 6523: Bug in RPackage >> systemSubPackages http://code.google.com/p/pharo/issues/detail?id=6523

Issue 6526: RPackage is not removing extensions from a removed class http://code.google.com/p/pharo/issues/detail?id=6526

-- Mariano http://marianopeck.wordpress.com

...

...

Shouldn't Jenkins help us here?

We have about three options: 1) use Jenkins for updates checking - maybe it was the original plan but it doesn't work because we do not want to rollback updates

I think the original plan was to only use images that completed all the tests. Much like the VM builds.

2) create a fake update server and create Jenkins jobs that will check the updates and will not pass the dirty ones to mainstream

I guess that goes along with #1, we should be able to rollback changes easily. And yes, we should have a 2 level publish system, first internally for checking and only then publicly..

3) improve the Monkey do do more checks on fixes of issues - to run more tests etc. And do not allow fixes that do not pass.

the monkey should filter out as much as possible. since at that point we didn't change anything yet, so it's the ideal moment to fix things :) - do strict testing (of everything) - do some code critics (undeclared vars ...) - provide simple heuristics about the complexity of the fix