Smalltalk › Pharo › Pharo Smalltalk Developers

Running tests in MCMethodDefinitionTest

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

3 messages Options

Torsten Bergmann

Running tests in MCMethodDefinitionTest

Hi,

a little story: while working on cleaning up MCMethodDefinitionTest (see https://pharo.fogbugz.com/f/cases/20830)
I noticed that from time to time you have trouble when running the test methods in this class one after the other
several times. So sometimes it worked and sometimes not.

I wondered since the latest Pharo 7 build #70360 is definitely green on CI, so the tests seem to work.

After investigating I've noticed that in MCMethodDefinitionTest>>tearDown a package "FooBarBaz" which
is used during the tests is unregistered and (when you run another test from this suite right afterwards)
is not found anymore. Because in the test initialization it is not.

This is easily fixed with setting the package up in the #setUp method which is unregistered in #tearDown
and one can easily run the tests several times independently.

So if one had trouble in Pharo 6 and Pharo 7 with sporadic fails in Monticello tests this might have been
the cause.

The fix is part of PR https://github.com/pharo-project/pharo/pull/583

Bye
T.

Marcus Denker-4

Re: Running tests in MCMethodDefinitionTest

Very nice! I am reviewing the change now.

Thanks,

Marcus

> On 6 Dec 2017, at 23:42, Torsten Bergmann <[hidden email]> wrote:
>
> Hi,
>
> a little story: while working on cleaning up MCMethodDefinitionTest (see https://pharo.fogbugz.com/f/cases/20830)
> I noticed that from time to time you have trouble when running the test methods in this class one after the other
> several times. So sometimes it worked and sometimes not.
>
> I wondered since the latest Pharo 7 build #70360 is definitely green on CI, so the tests seem to work.
>
> After investigating I've noticed that in MCMethodDefinitionTest>>tearDown a package "FooBarBaz" which
> is used during the tests is unregistered and (when you run another test from this suite right afterwards)
> is not found anymore. Because in the test initialization it is not.
>
> This is easily fixed with setting the package up in the #setUp method which is unregistered in #tearDown
> and one can easily run the tests several times independently.
>
> So if one had trouble in Pharo 6 and Pharo 7 with sporadic fails in Monticello tests this might have been
> the cause.
>
> The fix is part of PR https://github.com/pharo-project/pharo/pull/583
>
> Bye
> T.
>
>

Nicolas Cellier

Re: Running tests in MCMethodDefinitionTest

Thanks, very good find!

I reviewed the PR yesterday, all good to my eyes.

I backported the fix to Squeak where I just experienced the random failure when debugging opensmalltalk-vm test failures
https://github.com/hpi-swa/smalltalkCI/issues/350

2017-12-07 13:53 GMT+01:00 Marcus Denker <[hidden email]>:

Very nice! I am reviewing the change now.

Thanks,

Marcus

> On 6 Dec 2017, at 23:42, Torsten Bergmann <[hidden email]> wrote:
>
> Hi,
>
> a little story: while working on cleaning up MCMethodDefinitionTest (see https://pharo.fogbugz.com/f/cases/20830)
> I noticed that from time to time you have trouble when running the test methods in this class one after the other
> several times. So sometimes it worked and sometimes not.
>
> I wondered since the latest Pharo 7 build #70360 is definitely green on CI, so the tests seem to work.
>
> After investigating I've noticed that in MCMethodDefinitionTest>>tearDown a package "FooBarBaz" which
> is used during the tests is unregistered and (when you run another test from this suite right afterwards)
> is not found anymore. Because in the test initialization it is not.
>
> This is easily fixed with setting the package up in the #setUp method which is unregistered in #tearDown
> and one can easily run the tests several times independently.
>
> So if one had trouble in Pharo 6 and Pharo 7 with sporadic fails in Monticello tests this might have been
> the cause.
>
> The fix is part of PR https://github.com/pharo-project/pharo/pull/583
>
> Bye
> T.
>
>