[squeak-dev] 6 tests to go... Win7 results

Levente Uzonyi leves at elte.hu
Wed Oct 24 14:14:22 UTC 2012

On Wed, 24 Oct 2012, Frank Shearar wrote:

> On 23 October 2012 21:25, Herbert König <herbertkoenig at gmx.net> wrote:
>> Hi,
>>> We're down to 6 failing unit tests on the Linux platform, one of which
>>> is an expected failure and another of which is a known network issue
>>> for the Linux platform.
>>> Would my kind Windows and OSX build volunteers please run builds
>>> against the latest trunk image?
>>> (
>> Yesterday (Oct 22) I could download build #17 and got a TrunkImage.image and
>> .changes. Not the usual zip including the VM.
> The VM comes from the repository at
> https://github.com/frankshearar/squeak-ci. In particular, you can get
> a copy directly from here:
> https://github.com/frankshearar/squeak-ci/zipball/master
> Sorry, I thought I'd only need to mention the image location.
>> This is update 12251 according to the system reporter. I disable IPV6
>> support via preferences and run all tests in the Cog VM I used for the
>> previous tests. See http://minus.com/luZyqW2PRhIrq.
>> I get 3222 run, 3178 passes, 18 expected failures, 5 failures, 21 errors, 0
>> unexpected passes, as if the network preference is ignored again.
>> See http://minus.com/lpibYAKEi3bw5
>> I did this twice, as I didn't want to believe it.
>> If I update this image to trunk, it loads some updates and again claims to
>> be 12251 version.
>> Tests like before result in 3219 run, 3195 passes, 18 expected failures, 5
>> failures, 1 errors, 0 unexpected passes. See:
>> http://minus.com/l17FSZY5yOnhH
>> I have no idea but I still have that image if someone wants it.
>> http://minus.com/mbam1m6GnSJJJH is the complete Folder with the screenshots
>> of all runs I published up to now.
> Thanks very much for running the tests, Herbert. I appreciate it!
> This is actually what I _expected_, since one of the 6 failing tests
> is a network issue on the Linux platform. That is, I expect
> NetworkTests.Kernel.SocketTest.testSendTimeout to _pass_ on Windows.

It randomly passes on Vista, but fails most of the time. The problem is 
that the test expects that if it fills up the send buffer of a socket, 
then all subsequent send requests will fail, but I think the OS somehow 
makes the send buffer larger, so while sending failed at a point in time, 
later it will work again.

> That error is news to me though: FileDirectoryTest.testRelativeNameIfAbsolut...
> Would you mind sending more details on that failed test? I'd like to
> see which assertion fails.

That's an "old" error on Windows. The method (#relativeNameIfAbsoluteFor:) 
expects that an absolute path begins with the path delimitar, which is 
obviously not true on windows (C:\... doesn't begin with \). There are 
several absolute path formats on windows, so I think the method has to be 
implementation in DosFileDirectory.


> Thanks again!
> frank
>> Cheers
>> Herbert

More information about the Squeak-dev mailing list