Smalltalk › Squeak › Squeak - Dev

4.5 -- Tests status looking good!

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

3 messages Options

Chris Muller-4

4.5 -- Tests status looking good!

I'm trying to get a picture of our state of quality (for release) in
terms of our tests.

Here are failing tests in Trunk (as debuggable expressions):

1. (CompiledMethodTest selector: #testMethodClass) debug.
2. (ExceptionTests selector: #testHandlerFromAction) debug.
3. (LocaleTest selector: #testLocaleChanged) debug.
4. (MCEnvironmentLoadTest selector: #testLoadIntoEnvironment) debug.
5. (MethodHighlightingTests selector: #testMethodHighlighting) debug.
6. (MultiByteFileStreamTest selector: #testLineEndingChunk) debug.
7. (PackageDependencyTest selector: #testMorphic) debug.
8. (PackageDependencyTest selector: #testSUnitGUI) debug.
9. (ReleaseTest selector: #testMethodsWithUnboundGlobals) debug.
10. (ReleaseTest selector: #testNoObsoleteClasses) debug.
11. (ReleaseTest selector: #testSwapMouseButtonsPreference) debug.
12. (ReleaseTest selector: #testUndeclared) debug.
13. (ST80PackageDependencyTest selector: #testST80) debug.
14. (SocketTest selector: #testLocalAddress) debug.
15. (SocketTest selector: #testRemoteAddress) debug.
16. (WeakSetInspectorTest selector: #testSymbolTableM6812) debug.

==========

* 1. No idea. Something about Obsolete classes. Help needed!
2. Old problem. Fails in 4.4.
3. Fails in 4.4 too.
* 4. "Environment has no binding for MCMockClassA". No idea.
* 5. Did method highlighting change? Is the test wrong or broken? Eliot?
6. Something Nicolas noticed which is a problem in 4.4 too. So,
should not hold up 4.5.
* 7 and 8: ToolBuilder-Tests goof. Frank fixing now.
* 9 and 10: Several Obsolete classes. What is an Obsolete class, how
to get rid of it, and why did I want it in the first place?
11. My personal pref, will fix before release.
* 12. More Obsolete classes. They're Undeclared.
* 13. Same as 7 and 8.
14 and 15: are issues with my local address being [127 0 1 1] vs.
[127 0 0 1]. Same failure happens in 4.4. Not a show-stopper.
16. This is an issue with Inspector. I can take this one!

* = must be fixed prior to release.

Chris: 11, 16
Colin: 4
Eliot: 5
Frank: 7, 8, 13
Volunteer: 1, 9, 10, 12

So, we are looking GOOD tests-wise for release. If Colin and Eliot
can fix 4 and 5, and we can get volunteers for those other 4 (which
seem to be related to each other), then I think we'll be almost ready!

Someone who knows about Undeclareds and Obsoletes, please step up and
help Squeak 4.5 become a reality!

Levente Uzonyi-2

Re: 4.5 -- Tests status looking good!

Here's how I see these:

1,9,10,12: sign of issues with Environments. I tried to fixed them a few
times, but there are many issues which affect each other, and other parts
of the system too. Maybe I should push my fixes one by one, and see what
else is broken. Another problem with Environments is that some parts are
not polished yet.
2: a nice challenge :)
3: incomplete integration of localization related stuff from EToys
4: lack of support for environments in MC
5: something is broken there
6: easy to fix if we can find out what we want to do
11: it's just your image and vm combo
14,15: issues about name lookup in the "new" network code. We're still
using the old network code because of this in production images.

Levente

On Sat, 14 Dec 2013, Chris Muller wrote:

> I'm trying to get a picture of our state of quality (for release) in
> terms of our tests.
>
> Here are failing tests in Trunk (as debuggable expressions):
>
> 1. (CompiledMethodTest selector: #testMethodClass) debug.
> 2. (ExceptionTests selector: #testHandlerFromAction) debug.
> 3. (LocaleTest selector: #testLocaleChanged) debug.
> 4. (MCEnvironmentLoadTest selector: #testLoadIntoEnvironment) debug.
> 5. (MethodHighlightingTests selector: #testMethodHighlighting) debug.
> 6. (MultiByteFileStreamTest selector: #testLineEndingChunk) debug.
> 7. (PackageDependencyTest selector: #testMorphic) debug.
> 8. (PackageDependencyTest selector: #testSUnitGUI) debug.
> 9. (ReleaseTest selector: #testMethodsWithUnboundGlobals) debug.
> 10. (ReleaseTest selector: #testNoObsoleteClasses) debug.
> 11. (ReleaseTest selector: #testSwapMouseButtonsPreference) debug.
> 12. (ReleaseTest selector: #testUndeclared) debug.
> 13. (ST80PackageDependencyTest selector: #testST80) debug.
> 14. (SocketTest selector: #testLocalAddress) debug.
> 15. (SocketTest selector: #testRemoteAddress) debug.
> 16. (WeakSetInspectorTest selector: #testSymbolTableM6812) debug.
>
> ==========
>
> * 1. No idea. Something about Obsolete classes. Help needed!
> 2. Old problem. Fails in 4.4.
> 3. Fails in 4.4 too.
> * 4. "Environment has no binding for MCMockClassA". No idea.
> * 5. Did method highlighting change? Is the test wrong or broken? Eliot?
> 6. Something Nicolas noticed which is a problem in 4.4 too. So,
> should not hold up 4.5.
> * 7 and 8: ToolBuilder-Tests goof. Frank fixing now.
> * 9 and 10: Several Obsolete classes. What is an Obsolete class, how
> to get rid of it, and why did I want it in the first place?
> 11. My personal pref, will fix before release.
> * 12. More Obsolete classes. They're Undeclared.
> * 13. Same as 7 and 8.
> 14 and 15: are issues with my local address being [127 0 1 1] vs.
> [127 0 0 1]. Same failure happens in 4.4. Not a show-stopper.
> 16. This is an issue with Inspector. I can take this one!
>
> * = must be fixed prior to release.
>
> Chris: 11, 16
> Colin: 4
> Eliot: 5
> Frank: 7, 8, 13
> Volunteer: 1, 9, 10, 12
>
> So, we are looking GOOD tests-wise for release. If Colin and Eliot
> can fix 4 and 5, and we can get volunteers for those other 4 (which
> seem to be related to each other), then I think we'll be almost ready!
>
> Someone who knows about Undeclareds and Obsoletes, please step up and
> help Squeak 4.5 become a reality!
>
>

Colin Putney-3

Re: 4.5 -- Tests status looking good!

On Sun, Dec 15, 2013 at 2:46 AM, Levente Uzonyi <[hidden email]> wrote:

Here's how I see these:

1,9,10,12: sign of issues with Environments. I tried to fixed them a few times, but there are many issues which affect each other, and other parts of the system too. Maybe I should push my fixes one by one, and see what else is broken. Another problem with Environments is that some parts are not polished yet.

Hi everybody,

A while back Levent mentioned that we ought to initialize the references dictionary in an Environment eagerly rather than lazily. (http://lists.squeakfoundation.org/pipermail/squeak-dev/2013-November/174469.html)

I think this is a good idea, because it makes Environments more similar to the existing single-namespace system. It's possible to reflect on the visible bindings in a single place, rather than have them magically appear when you try to use them. However, it means that we'll need to modify the references dictionary in response to changes in other environments in order to keep it up today. That implies hooking into SystemChangeNotifications, and modifying them to include environment information.

I've started to work on this change, and I hope to fix the above issues along the way. I've also run into some semantic issues that need some clarification, which I'll bring up in separate messages.

Colin