Supported platforms = ? (was Re: [squeak-dev] The Trunk: NetworkTests-ul.32.mcz)

> On Tue, Aug 14, 2012 at 07:11:26AM +0100, Frank Shearar wrote:
>> On 13 August 2012 21:55, Levente Uzonyi <[hidden email]> wrote:
>> > On Mon, 13 Aug 2012, Frank Shearar wrote:
>> >
>> >> On 13 August 2012 16:37,  <[hidden email]> wrote:
>> >>>
>> >>> Levente Uzonyi uploaded a new version of NetworkTests to project The
>> >>> Trunk:
>> >>> http://source.squeak.org/trunk/NetworkTests-ul.32.mcz
>> >>>
>> >>> ==================== Summary ====================
>> >>>
>> >>> Name: NetworkTests-ul.32
>> >>> Author: ul
>> >>> Time: 13 August 2012, 5:36:58.512 pm
>> >>> UUID: 890b13a3-3ca4-b840-ae6c-f2a252618455
>> >>> Ancestors: NetworkTests-dtl.31
>> >>>
>> >>> Added some tests for Socket >> #peerName.
>> >>>
>> >>> =============== Diff against NetworkTests-dtl.31 ===============
>> >>>
>> >>> Item was added:
>> >>> + ----- Method: SocketTest>>testPeerName (in category 'tests') -----
>> >>> + testPeerName
>> >>> +
>> >>> +       self shouldnt: [ Socket new peerName ] raise: Error.
>> >>> +       self testServerAccept.
>> >>> +       self shouldnt: [ listenerSocket peerName ] raise: Error.
>> >>> +       self shouldnt: [ clientSocket peerName ] raise: Error.
>> >>> +       self shouldnt: [ serverSocket peerName ] raise: Error!
>> >>
>> >>
>> >> This test fails on the build server:
>> >>
>> >>
>> >> http://squeakci.org/job/SqueakTrunk/33/testReport/NetworkTests.Kernel/SocketTest/testPeerName/
>> >>
>> >> Less than helpful debug stuff there - "Assertion failed" - but there you
>> >> go.
>> >>
>> >> I would imagine it's yet another platform specific issue?
>> >
>> >
>> > Does the VM have the new network primitives?
>>
>> It's Cog r2652, so it ought to for sufficiently new meanings of "new".
>>
>
> Cog does not provide the new network primitives. You can verify whether
> your VM has the primitives by evaluating "NetNameResolver useOldNetwork".
>
> This may be an issue in Socket class>>remoteSocketAddress that is
> exposed when running with the new network support. The test attempts
> to do "Socket new peerName" which does not work (but should it???).

> On 14 August 2012 13:34, David T. Lewis <[hidden email]> wrote:
> >
> > Cog does not provide the new network primitives. You can verify whether
> > your VM has the primitives by evaluating "NetNameResolver useOldNetwork".
> >
> > This may be an issue in Socket class>>remoteSocketAddress that is
> > exposed when running with the new network support. The test attempts
> > to do "Socket new peerName" which does not work (but should it???).
>
> Right now the CI environment _only_ runs Cog + Linux. At the very
> least we should test on Windows and OSX too. But the reason for the
> change in subject is that it sounds like we also need to test several
> _VMs_ + platform?

> On Tue, Aug 14, 2012 at 02:17:40PM +0100, Frank Shearar wrote:
>> On 14 August 2012 13:34, David T. Lewis <[hidden email]> wrote:
>> >
>> > Cog does not provide the new network primitives. You can verify whether
>> > your VM has the primitives by evaluating "NetNameResolver useOldNetwork".
>> >
>> > This may be an issue in Socket class>>remoteSocketAddress that is
>> > exposed when running with the new network support. The test attempts
>> > to do "Socket new peerName" which does not work (but should it???).
>>
>> Right now the CI environment _only_ runs Cog + Linux. At the very
>> least we should test on Windows and OSX too. But the reason for the
>> change in subject is that it sounds like we also need to test several
>> _VMs_ + platform?
>
> In a perfect world we would test the image on a range of operating
> systems and VMs, including out of date VMs, old operating systems,
> old computers, 32-bit and 64-bit VMs, etc, etc. It would be worthwhile
> to test both Cog and SqueakVM on Linux, but extending the range beyond
> that might be a lot of work.