[icedtea-web] RFC: add unit tests for the jnlp parser

[akurtakov]

On Mon, Apr 4, 2011 at 3:57 PM, Jiri Vanek <jvanek at redhat.com> wrote:
>
>
> -------- Original Message --------
> Subject: Re: [icedtea-web] RFC: add unit tests for the jnlp parser
> Date: Mon, 04 Apr 2011 13:45:15 +0200
> From: Jiri Vanek <jvanek at redhat.com>
> To: distro-pkg-dev at openjdk.java.net
>
> On 04/01/2011 11:54 PM, Omair Majid wrote:
>>
>> On 03/31/2011 04:56 PM, Dr Andrew John Hughes wrote:
>>>
>>> On 15:30 Thu 31 Mar , Omair Majid wrote:
>>>>
>>>> On 03/31/2011 02:26 PM, Dr Andrew John Hughes wrote:
>>>>>
>>>>> On 13:40 Thu 31 Mar , Omair Majid wrote:
>>>>>>
>>>>>> On 03/30/2011 08:21 PM, Dr Andrew John Hughes wrote:
>>>>>>>
>>>>>>> It does need documenting in README, but I'm happy for that to be a
>>>>>>> separate patch.
>>>>>>
>>>>>> Patch attached. Does that look okay?
>>>>>>
>>>>>
>>>>> It's a bit brief.
>>>>>
>>>>> * What about documenting the --with-rhino/junit options?
>>>>> * What about something specific about the JUnit tests?
>>>>
>>>> Okay, how about the new patch?
>>>>
>>> Still nothing on the JUnit stuff, which was the main reason I thought
>>> we were
>>> patching the docs. The custom output and the reasoning for it should
>>> be noted
>>> IMHO.
>>>
>>
>> Ah, _now_ I get it. I thought by JUnit, you were referring to JUnit
>> itself, not our frontend/formatter for it.
>>
>> I thought my documentation of that in tests/junit-runner/README was
>> sufficient. Anyway, I have updated the README for that too. Any other
>> suggestions?
>>
>> Cheers,
>> Omair
>
> Today I have tested and  walked through junit test for netx. Compraed to
> TestNg I consider it as big step  back. I still do not see, why 3jars of
> easily downloadable dependences

Let me add a distro POV. That's the problem for distributions "easily
downloadable dependencies"  but most distributions do build from
source and usually "easily downloadable" doesn't mean "easily
buildable" . TestNG is a clasical example - it needs guice, jcommon,
bsh, snakeyaml .... + all of their dependencies. This limits distros
to ship older versions of such libs. E.g. in Fedora we ship testng
5.11.3 (Debian ships 5.11.x version too) with quite minimal chances
for getting newer testng anytime soon. On the contra side junit is
buildable with plain ant and just hamcrest as a dependency(also
buildable with plain ant). If one choose to use testng he will have to
limit his usage to quite old version of testng (possibly unsupported
now ??) or put a bit of work in the major distros to put the needed
packages in place. I agree that one can say that tests should be run
by developers but if you don't run them on the bits your users will
use you will lose  a lot of their value.

P.S. This mail is not saying anything about testng vs. junit on the
technical site. It's just pointing one major point that
developers(especially Java one) use to forget when considering their
dependencies - keeping their dependencies in a state which won't force
downstreams to hack/patch/dumb their  build systems resulting in
shipping software that is not build in exactly the same way upstream
does.

Regards,
Alex


> (can be handled as drops), one
> nicely-readable xml and annotations configuration and few lines of
> make.am/configure and extremely good looking and informing output and
> clear license was abandoned  for  one jar of dependence, mess makefile,
> nearly no  configuration posibility, and custom java-class luncher
> (extending such a complex thing can always bring troubles  and much
> confusion inside), maybe confusing license and its quite strange
> application (I'm not  deep enough in it).
>
> TestNG - advantages - output, configuration(this is huge problem in
> JUnit), small makefile integration, license, documentation
>        disadvantages - 2 more jars as dependencies
> JUnit - advantages - no dependences
>        disadvantages - output, custom java-class luncher, no configuration,
> license, to much of makefile in (find *class... :-/// )
>
>
>
> Google can be my witness;)
> http://www.google.com/search?q=junit+testng+comparison
>
>
> Basicly I do not agree with Andrews arguments against TestNG. Eg. Deepak
> agreed with them. Omair found them sufficient, And I'm finding them much
> better.
>
>
> Sorry for post-commit reaction but I have overlooked swap testng->junit
> in distro emails.
>
>
> Regards J.
>
>