[rfc][icedtea-web] Unittest regression fix

[Jiri Vanek]

On 02/10/2014 03:47 PM, Andrew Azores wrote:
> Hi,
>
> Jiri noticed that a recent patch caused 10 unit test failures, due to the classloader bailing out on the condition that codebase loading was disabled and there appeared to be no JARs to load from. This conclusion was apparently reached too quickly and was caused because I had assumed this information would already be ready, and so had moved this logic earlier in the Launcher/ClassLoader cycle. Moving the check back to where it was, ie later on, solves the problem.
>
> ChangeLog:
> Partial revert of 7933143a1286, refactoring to move codebase-loading-enabling
> logic out of Launcher and into JNLPClassLoader.
> * netx/net/sourceforge/jnlp/Launcher.java: (createApplet, createAppletObject):
> handle enableCodebase again
> * netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java: (enableCodebase):
> re-added, codebase enabling logic moved back out into Launcher
>
> Thanks,
>
Hi. This changes looks ok.
However:
  - this regression was cought by unittests. You as developer are supposed to check *each* (your) changeset with unittest (there is no excuse for not do it) and at least with some subset of reproducers. Ideally to let them run over night all.
  - this fix have got really broken. I think that Except original patch there is already patch to original one, and patch to patch(this one).  I'm really unhappy from it. I do not insists, But I would recommend to revert previous two, mingle with this one, and revisite as one complex patch.

Anyway this one is good to go to have head a bit stable again.

However I would strongly encourage you to revisit the patch as whole.


J.