On Wed, 22 Dec 2010, Chris Muller wrote:

> Here are the results of running the tests in a 10802 image:
>
>   2828 run, 2810 passes, 6 expected failures, 12 failures, 0 errors,
> 0 unexpected passes
>
> The failed tests:
>
> BecomeTest>>#testBecomeIdentityHash

It has some chance to fail, but it fails more often on Cog. I think the 
last assert should be removed.

> DecompilerTests>>#testDecompileLoopWithMovingLimit

http://bugs.squeak.org/view.php?id=7093

> DecompilerTests>>#testDecompilerInClassesMAtoMM
> DecompilerTests>>#testDecompilerInClassesSNtoSZ

These two have possible fixes in the Inbox.

> FileStreamTest>>#testPositionPastEndIsAtEnd

This is with an older version of CogVM. This is fixed in the svn repo.

> MCDictionaryRepositoryTest>>#testCreationMethods
> MCDirectoryRepositoryTest>>#testCreationMethods

These don't fail on windows.

> MCPackageTest>>#testUnload

This fails randomly. If you run it alone, then it doesn't. If we could 
save the context of the failing test and open the debugger with that 
context later, then it would be a lot easier to fix such issues.

> PackageDependencyTest>>#testMorphic
> PackageDependencyTest>>#testSUnitGUI

Two dependencies to be resolved.

> TestURI>>#testDirWithHash

A possible fix is in the Inbox.

>
> I've looked at two of them and.. I have no idea.  Especially curious
> for me are the FileStreamTest and BecomeTest.  Strange, the code
> hasn't changed since 4.1...
>
> The possible outcomes for Squeak 4.2 are:
>
>  a) I override #shouldPass to mark these tests as "expected failures".
>  b) Volunteers fix the remaining tests within the next couple of days.
>  c) Release 4.2 with yellow tests.
>  d) Delay the 4.2 release and hope volunteers fix the remaining tests.
>  e) ???
>
> My top preference is (b) of course, but if that does not happen, then
> (c) or (a).

I think we should wait for the new VMs, before releasing the new images.


Levente

>
>  - Chris
>
>