On 28 October 2012 13:04, David T. Lewis <lewis at mail.msen.com> wrote:
> On Sun, Oct 28, 2012 at 12:39:34AM -0400, David T. Lewis wrote:
>> On Sat, Oct 27, 2012 at 11:41:15PM +0100, Frank Shearar wrote:
>> > On 27 October 2012 23:37,  <commits at source.squeak.org> wrote:
>> > > Frank Shearar uploaded a new version of KernelTests to project The Inbox:
>> > > http://source.squeak.org/inbox/KernelTests-fbs.235.mcz
>> > >
>> > > ==================== Summary ====================
>> > >
>> > > Name: KernelTests-fbs.235
>> > > Author: fbs
>> > > Time: 27 October 2012, 11:37:53.223 pm
>> > > UUID: ab0d84db-c1bc-40d0-bd90-786dcd6989b5
>> > > Ancestors: KernelTests-ul.234
>> > >
>> > > Mark this test as an expected failure for the 4.4 release. Unmark it at the start of the 4.5 cycle!
>> > >
>> > > =============== Diff against KernelTests-ul.234 ===============
>> > >
>> > > Item was added:
>> > > + ----- Method: LargeNegativeIntegerTest>>expectedFailures (in category 'as yet unclassified') -----
>> > > + expectedFailures
>> > > +       ^ #(testMinimumNegativeIntegerArithmetic).!
>> >
>> > IIRC this test can only pass with suitable VM changes being made.
>> > http://bugs.squeak.org/view.php?id=7705 mentions that the fix is in
>> > VMMaker-dtl.286, so does this mean that with the interpreter VM this
>> > test will pass, but with Cog (the VM we use in CI tests) the test will
>> > fail?
>>
>> Yes, that's right. It will fail with older interpreter VMs also.
>>
>
> Frank,
>
> If you want to try running the test suite with an interpreter VM, I
> installed a locally built VM on box3.squeak.org that can be used to
> drive the tests. This is installed as /usr/local/bin/squeak, and can
> be used for running both 32-bit and 64-bit Squeak images.
>
> This VM is built from latest VMMaker and squeakvm.org sources using
> the ~lewis/VMUnixBuild/VMUnixBuild.st script (attached), which now seems
> to work fairly reliably. It is *not* suitable for general distribution,
> so I would appreciate if we can all be careful to refer to it as a local
> build, and not as "the standard VM" or any words to that effect.

Thanks Dave, I'll be able to pick this up in a few hours, once my boys
have gone to bed. I'll add a new script to the squeak-ci repo that
refers to the local build. (It's nearly time to refactor the build
scripts, too.)

> Note, the latest standard interpreter VM (*) is not working on our
> box3.squeak.org server due to a Linux glibc compatibility issue
> (/lib/libc.so.6: version `GLIBC_2.15' not found), hence the need for
> a local build.

Yes, I ran into this the last time I tried to set up an interpreter VM
build, thanks to the awesomeness of CentOS.

frank

> (*) http://squeakvm.org/unix/release/Squeak-4.10.2.2614-linux_i386.tar.gz
>
> Dave
>
>
>
>