Smalltalk › Squeak › Squeak - Dev

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

14 messages Options

Hannes Hirzel

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

I redid the test now with adding an Inbound and Outbound rule in the
Windows 7 (Ultimate edition) advance firewall settings dialog.

http://i.minus.com/ibwcCvF6neD876.png

However the number of tests failing and the number of errors is still high.

http://i.minus.com/imTLFlio5Huyk.png
17 expected failures, 21 failures, 238 errors.

The reason for this might be that this time I did not have any
internet connection.

Frank wrote that he will post a new download link for a zip file with
everything included (cogVM for Windows, image and changes files,
sources file).

I will then redo the test.

--Hannes

Levente Uzonyi-2

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

On Thu, 6 Sep 2012, H. Hirzel wrote:

> I redid the test now with adding an Inbound and Outbound rule in the
> Windows 7 (Ultimate edition) advance firewall settings dialog.
>
> http://i.minus.com/ibwcCvF6neD876.png
>
> However the number of tests failing and the number of errors is still high.
>
> http://i.minus.com/imTLFlio5Huyk.png
> 17 expected failures, 21 failures, 238 errors.
>
> The reason for this might be that this time I did not have any
> internet connection.

It seems to me that the image has no write access to its directory.

Levente

>
> Frank wrote that he will post a new download link for a zip file with
> everything included (cogVM for Windows, image and changes files,
> sources file).
>
> I will then redo the test.
>
> --Hannes
>
>

Frank Shearar-3

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

In reply to this post by Hannes Hirzel

On 6 September 2012 15:12, H. Hirzel <[hidden email]> wrote:

Sorry for taking so long. It should be in the zip now.

frank

> I will then redo the test.
>
> --Hannes
>

Hannes Hirzel

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Levente,

thank you for the note about missing write access for croquet.exe.
I assume that it is how it is supposed to be with Windows 7
http://en.wikipedia.org/wiki/User_Account_Control

I wonder if I can give TestRunner another default directory to put the
test data into.

Frank,

you mean this one
https://github.com/frankshearar/squeak-ci/zipball/master
?

--Hannes

On 9/6/12, Frank Shearar <[hidden email]> wrote:

> On 6 September 2012 15:12, H. Hirzel <[hidden email]> wrote:
>> I redid the test now with adding an Inbound and Outbound rule in the
>> Windows 7 (Ultimate edition) advance firewall settings dialog.
>>
>> http://i.minus.com/ibwcCvF6neD876.png
>>
>> However the number of tests failing and the number of errors is still
>> high.
>>
>> http://i.minus.com/imTLFlio5Huyk.png
>> 17 expected failures, 21 failures, 238 errors.
>>
>> The reason for this might be that this time I did not have any
>> internet connection.
>>
>> Frank wrote that he will post a new download link for a zip file with
>> everything included (cogVM for Windows, image and changes files,
>> sources file).
>
> Sorry for taking so long. It should be in the zip now.
>
> frank
>
>> I will then redo the test.
>>
>> --Hannes
>>
>
>

Frank Shearar-3

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

On 6 September 2012 17:37, H. Hirzel <[hidden email]> wrote:

> Levente,
>
> thank you for the note about missing write access for croquet.exe.
> I assume that it is how it is supposed to be with Windows 7
> http://en.wikipedia.org/wiki/User_Account_Control
>
> I wonder if I can give TestRunner another default directory to put the
> test data into.
>
>
> Frank,
>
> you mean this one
> https://github.com/frankshearar/squeak-ci/zipball/master
> ?

Yes: I would imagine that that URL would give you a _new_ zip based
off the latest commit to that repository.

frank

> --Hannes
>
> On 9/6/12, Frank Shearar <[hidden email]> wrote:
>> On 6 September 2012 15:12, H. Hirzel <[hidden email]> wrote:
>>> I redid the test now with adding an Inbound and Outbound rule in the
>>> Windows 7 (Ultimate edition) advance firewall settings dialog.
>>>
>>> http://i.minus.com/ibwcCvF6neD876.png
>>>
>>> However the number of tests failing and the number of errors is still
>>> high.
>>>
>>> http://i.minus.com/imTLFlio5Huyk.png
>>> 17 expected failures, 21 failures, 238 errors.
>>>
>>> The reason for this might be that this time I did not have any
>>> internet connection.
>>>
>>> Frank wrote that he will post a new download link for a zip file with
>>> everything included (cogVM for Windows, image and changes files,
>>> sources file).
>>
>> Sorry for taking so long. It should be in the zip now.
>>
>> frank
>>
>>> I will then redo the test.
>>>
>>> --Hannes
>>>
>>
>>
>

Hannes Hirzel

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Thank you for the URL and I downloaded the zip.

However as of now I am stuck with a problem that Windows 7 refuses to
remove the read-only flag on the squeak folder though I tell it to do
so (Explorer in Admin mode).
I think I'll need to apply all updates to Windows 7.

--Hannes

On 9/6/12, Frank Shearar <[hidden email]> wrote:

> On 6 September 2012 17:37, H. Hirzel <[hidden email]> wrote:
>> Levente,
>>
>> thank you for the note about missing write access for croquet.exe.
>> I assume that it is how it is supposed to be with Windows 7
>> http://en.wikipedia.org/wiki/User_Account_Control
>>
>> I wonder if I can give TestRunner another default directory to put the
>> test data into.
>>
>>
>> Frank,
>>
>> you mean this one
>> https://github.com/frankshearar/squeak-ci/zipball/master
>> ?
>
> Yes: I would imagine that that URL would give you a _new_ zip based
> off the latest commit to that repository.
>
> frank
>
>> --Hannes
>>
>> On 9/6/12, Frank Shearar <[hidden email]> wrote:
>>> On 6 September 2012 15:12, H. Hirzel <[hidden email]> wrote:
>>>> I redid the test now with adding an Inbound and Outbound rule in the
>>>> Windows 7 (Ultimate edition) advance firewall settings dialog.
>>>>
>>>> http://i.minus.com/ibwcCvF6neD876.png
>>>>
>>>> However the number of tests failing and the number of errors is still
>>>> high.
>>>>
>>>> http://i.minus.com/imTLFlio5Huyk.png
>>>> 17 expected failures, 21 failures, 238 errors.
>>>>
>>>> The reason for this might be that this time I did not have any
>>>> internet connection.
>>>>
>>>> Frank wrote that he will post a new download link for a zip file with
>>>> everything included (cogVM for Windows, image and changes files,
>>>> sources file).
>>>
>>> Sorry for taking so long. It should be in the zip now.
>>>
>>> frank
>>>
>>>> I will then redo the test.
>>>>
>>>> --Hannes
>>>>
>>>
>>>
>>
>
>

Herbert König

Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Hi,

Am 06.09.2012 23:49, schrieb H. Hirzel:
> Thank you for the URL and I downloaded the zip.
>
>
There's several images in that zip
Squeak4.4-12053.image, seems quite old, maybe it should not show up in
the zip. Same with the Squeak4.3.image
My results from this one:
3177 run 3148 passes, 16 expected failures, 12 failures, 1 errors, 0
unexpected passes
Everything goes smooth, no wait.

Squeak4.4-trunk.image
again shows that wait, this time after the first test and:
3204 run 3157 passes, 17 expected failures, 10 failures, 20 errors, 0
unexpected passes

Croquet has permissions for UDP and TCP on all ports. So I gave
permission for all protocols and all ports, same result.

Next I disabled WiFi, no wired connection available and disabled the
firewall. Again the same result.

Copied the Folder to an XP virtual machine and got:
3201 run, 3173 passes, 17 expected failures, 10 failures, 1 errors, 0
unexpected passes
SocketTest>>testPeername failed as the only socket test
http://minus.com/lbloGF7LskM0a7

Cheers,

Herbert

Frank Shearar-3

Re: Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

On 7 September 2012 14:16, Herbert König <[hidden email]> wrote:

> Hi,
>
> Am 06.09.2012 23:49, schrieb H. Hirzel:
>>
>> Thank you for the URL and I downloaded the zip.
>>
>>
> There's several images in that zip
> Squeak4.4-12053.image, seems quite old, maybe it should not show up in the
> zip. Same with the Squeak4.3.image

It's a zip of the working copy of the repository, so it will contain
several things that are _not_ in the zip produced by Jenkins. They're
the base images used by the scripts to build a fully up-to-date image.
Since you're interested in the zip mainly for the Cog VMs, I'd suggest
just ignoring the rest of the zip for the moment.

You _could_ download the appropriate Cog VM directly, but pulling them
from a freshly downloaded zip ensures your tests run on the same VM as
the buildserver does (modulo platform).

frank

> My results from this one:
> 3177 run 3148 passes, 16 expected failures, 12 failures, 1 errors, 0
> unexpected passes
> Everything goes smooth, no wait.
>
> Squeak4.4-trunk.image
> again shows that wait, this time after the first test and:
> 3204 run 3157 passes, 17 expected failures, 10 failures, 20 errors, 0
> unexpected passes
>
> Croquet has permissions for UDP and TCP on all ports. So I gave permission
> for all protocols and all ports, same result.
>
> Next I disabled WiFi, no wired connection available and disabled the
> firewall. Again the same result.
>
> Copied the Folder to an XP virtual machine and got:
> 3201 run, 3173 passes, 17 expected failures, 10 failures, 1 errors, 0
> unexpected passes
> SocketTest>>testPeername failed as the only socket test
> http://minus.com/lbloGF7LskM0a7
>
> Cheers,
>
> Herbert
>
>

Herbert König

Re: Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Hi,

It's a zip of the working copy of the repository, so it will contain
several things that are _not_ in the zip produced by Jenkins. They're
the base images used by the scripts to build a fully up-to-date image.
Since you're interested in the zip mainly for the Cog VMs, I'd suggest
just ignoring the rest of the zip for the moment.

As there was no new versions I just upgraded the trunk image in the zip to 12211 and reran the tests.
I tried to debug the socket issue.

ArbitraryObjectSocketTestcase>>setUp calls NetNameResolver>>localHostAddress which returns:
fe80::64bb:69f6:38d0:b4bf%20(AldiLaptop1),0(0)

As I have no knowledge about IPV6: is this a valid address? I'm used to getting a ByteArray from this.

Higher up the stack in
Socket>>waitForConnectionFor: timeout ifTimedOut: timeoutBlock ifRefused: refusedBlock
the refusedBlock is evaluated which hints that the above address is ok but it's still a firewall issue.

Picture below is from the TCP firewall rule maybe I have to grant rights for some of the IPV6 options?
Anyway, we'd have to document this because my windows is "Out of the box" and I use the standard interactive way to punch holes into the firewall like may users will do.

Any knowledgeable "Windowers" around?

Thanks,

Herbert

Herbert König

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

In reply to this post by Hannes Hirzel

Hi,

just tested #12176 with no changes.
3204 run, 3157 passes, 17 expected failures, 10 failures, 20 errors, 0
unexpected passes
Cheers,

Herbert

Herbert König

Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Am 09.09.2012 08:06, schrieb Herbert König:
> Hi,
>
> just tested #12176 with no changes.
> 3204 run, 3157 passes, 17 expected failures, 10 failures, 20 errors, 0
> unexpected passes

Hi,

same with #12176. Had a look at Pharo, there
NetNameResolver>>localHostAddress returns the old byte Array (1.4 and
2.0) so no help from there too.

Cheers

Herbert

Levente Uzonyi-2

Re: Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

In reply to this post by Herbert König

On Sat, 8 Sep 2012, Herbert König wrote:

> Hi,
>
>> It's a zip of the working copy of the repository, so it will contain
>> several things that are _not_ in the zip produced by Jenkins. They're
>> the base images used by the scripts to build a fully up-to-date image.
>> Since you're interested in the zip mainly for the Cog VMs, I'd suggest
>> just ignoring the rest of the zip for the moment.
>>
>>
> As there was no new versions I just upgraded the trunk image in the zip to
> 12211 and reran the tests.
> I tried to debug the socket issue.
>
> ArbitraryObjectSocketTestcase>>setUp calls NetNameResolver>>localHostAddress
> which returns:
> fe80::64bb:69f6:38d0:b4bf%20(AldiLaptop1),0(0)
>
> As I have no knowledge about IPV6: is this a valid address? I'm used to
> getting a ByteArray from this.

This is a link-local address, which was assigned by the OS automatically.
Same as the 169.254.x.x in IPv4.

>
> Higher up the stack in
> Socket>>waitForConnectionFor: timeout ifTimedOut: timeoutBlock ifRefused:
> refusedBlock
> the refusedBlock is evaluated which hints that the above address is ok but
> it's still a firewall issue.
>
> Picture below is from the TCP firewall rule maybe I have to grant rights for
> some of the IPV6 options?

Yes, that should do the trick, please try it.

Levente

> Anyway, we'd have to document this because my windows is "Out of the box" and
> I use the standard interactive way to punch holes into the firewall like may
> users will do.
>
> Any knowledgeable "Windowers" around?
>
> Thanks,
>
> Herbert
>
>
>

Herbert König

Re: Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

Hi,
Am 10.09.2012 16:47, schrieb Levente Uzonyi:
>>
>> Picture below is from the TCP firewall rule maybe I have to grant
>> rights for some of the IPV6 options?
>
>
> Yes, that should do the trick, please try it.
>
I tried each of them without success.
Then I switched off WiFi (no cable network here), disabled the firewall
altogether and just tried to run SocketTest>>testClientConnect.

I get a "connection refused" at once (not after the timeout which I
raised to 20 seconds).
Debugging brings me down to
Socket>>primSocket: socketHandle connectTo: aSocketAddress.
with socketHandle being
#[249 80 153 11 0 0 0 0 24 23 194 5]
and aSocketAddress is
fe80::3883:fb18:2d77:313d%30(AldiLaptop1),42324(42324)

Does the internet router have to support IPV6 for the tests to work?
Unlikely for the loopback device.

Cheers,

Herbert

Levente Uzonyi-2

Re: Windows 7 tests behaviour depending on image was: Re: [TestRunner] Windows 7 tests -- no longer weird behaviour

On Mon, 10 Sep 2012, Herbert König wrote:

> Hi,
> Am 10.09.2012 16:47, schrieb Levente Uzonyi:
>>>
>>> Picture below is from the TCP firewall rule maybe I have to grant rights
>>> for some of the IPV6 options?
>>
>>
>> Yes, that should do the trick, please try it.
>>
> I tried each of them without success.
> Then I switched off WiFi (no cable network here), disabled the firewall
> altogether and just tried to run SocketTest>>testClientConnect.
>
> I get a "connection refused" at once (not after the timeout which I raised to
> 20 seconds).
> Debugging brings me down to
> Socket>>primSocket: socketHandle connectTo: aSocketAddress.
> with socketHandle being
> #[249 80 153 11 0 0 0 0 24 23 194 5]
> and aSocketAddress is
> fe80::3883:fb18:2d77:313d%30(AldiLaptop1),42324(42324)
>
> Does the internet router have to support IPV6 for the tests to work? Unlikely
> for the loopback device.

No, they should just work.

Levente

>
> Cheers,
>
> Herbert
>
>
>