[RFC][icedtea-web] Do not allow searching of jar manifest classpath if jnlp_href is being used

[Danesh Dadachanji]

On 15/08/12 05:24 AM, Jiri Vanek wrote:
> On 08/14/2012 09:15 PM, Danesh Dadachanji wrote:
>> On 14/08/12 07:44 AM, Jiri Vanek wrote:
>>> On 07/09/2012 04:51 PM, Danesh Dadachanji wrote:
>>>> With the many improvements in the test engine, I have been able to write up the tests for this
>>>> finally! Many thanks to Jiri for all
>>>> the features added to the test engine!
>>>>
>>>> ChangeLog:
>>>> +2012-07-09 Danesh Dadachanji <ddadacha at redhat.com>
>>>> +
>>>> + Classpaths in jars' manifests are only considered when the applet is run
>>>> + without using jnlp_href and a JNLP file.
>>>> + * netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java (activateJars):
>>>> + Add conditional check for use of jnlp_href.
>>>> + * tests/reproducers/signed/Classpath.Manifest.Test.Helper/srcs/CheckForClasspath.java:
>>>> + Applet whose jar is stored in a subdir under the test engine server.
>>>> + * tests/reproducers/signed/ClasspathManifestTest/resources/ClasspathManifestAppletTest.html:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/resources/ClasspathManifestAppletTest.jnlp:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/resources/ClasspathManifestApplicationTest.jnlp:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/resources/ClasspathManifestJNLPHrefTest.html:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/srcs/ClasspathManifest.java:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/srcs/META-INF/MANIFEST.MF:
>>>> + * tests/reproducers/signed/ClasspathManifestTest/testcases/ClasspathManifestTest.java:
>>>> + Test if manifest entry is searched for classpath only when in the plugin
>>>> + is run without using jnlp_href.
>>>> +
>>>>
>>>> I've updated my changeset (first attachment), I had forgotten a '!' in the if-statement there. Is
>>>> this one okay for HEAD and to
>>>> backport?
>>>>
>>>> The above scenarios that are supposed to fail (applets run by some JNLP file) do not fail in HEAD
>>>> when run locally. I ran into some
>>>> issues with running the above tests locally though, the X11 display is not found. I've attached a
>>>> second patch that will add said
>>>> tests. It's missing one, the passing browser test.
>>>>
>>>
>>> I have not noticed those. Everything worked fine on head.
>>
>> Weird, I tried again on my local machine and it complains but it works on another one. However, this
>> raises another issue. This fix only works for remote application launchers, classpath is still
>> searched when you run locally. E.g. file:///some/path/to/test.jnlp will search classpath. I've
>> attached my latest patch just for records.
>
> Yap. I remember that. Thats why there was originally added and with @KnownToFail - that there is issue, but will not be fixed soon, as
> minor issue with local files only.

Apparently my brain wasn't functioning enough to remember that. :/ Thanks for pointing that out and for all the reviews! I have pushed 
the change[1] and have also updated the wiki noting it[2]. I did not close PR975 just yet, I'd rather it be closed once the locally run 
files deny manifest search as well.

Regards,
Danesh

[1] http://icedtea.classpath.org/hg/icedtea-web/rev/fafa18b429f4
[2] http://icedtea.classpath.org/wiki/IcedTea-Web#Classpath_Usage_in_Manifests