[rfc][icedtea-web] try to properly(?) react on java/j2se version

[Jiri Vanek]

On 02/02/2015 08:35 PM, Jie Kang wrote:
>
>
> ----- Original Message -----
>> hi!
>>
>> I was recently acknowledged with ITW misbehaving when interpretting
>> java/j2se's version attribute.
>> When they "dont match" then itw simply continues(???)
>>
>> The useces I was informed about, was, that the application was known not to
>> run on jdk7
>>
>> So they included `version="1.6.0"` and itw happily runs without any warning.
>>
>>
>> I'm adding reaction - in strict mode dye, in non strict mode popup window.
>>
>> Actually  - I'm absolutely unsure what to do with this case...
>
> Hello,
>
>
> Is there a website I can go for testing this? Or could you create a reproducer for this?

one BAD example is
http://argouml.tigris.org/files/documents/4/383/ArgoUML-stable.jnlp

the behavior of this appis exactly why not to do what i did.
  - forgoten jnlp app
  - rewuires exactly 1.4.whatever
  => unwonted dialog rise up


>
>
> +    public static class JreVersion extends Version{
>
> I'd strongly prefer this to be in it's own file along with it's own tests.

hmhhm.. why? :(( It is so single case, that I dont wont to puboic it more.
>
> As well can you add some comments to the differences between JreVersion and Version. When do we use Version instead of JreVersion and when do we use JreVersion instead of Version? It isn't 100% clear to me at the moment;;
>
> +                    throw new RuntimeException("Strict run is  deffined, and your JRE - " + getJreVersion() + " - dont match requested JRE - " + v);
>
> s/deffined/defined
>
> Should this string be localized?
nope - localised exceptions are root of another evil.
But I will localize the message which goes to dialogue - if some dialogue appears at all.
>
> +        //this fails, do we need workaround?
> +        //Assert.assertTrue(new Version("1.5-").matches("1.4"));
>
> I think rather than workaround, a patch to the Version class would be appropriate.

I will change it to:
  +        //this fails, do we need patch it?
  +        //Assert.assertTrue(new Version("1.5-").matches("1.4"));
>
>

My concerns are that - is nowhere speciifed.. and dont give much sense...



A way - I'm thinking of two changes

for this patch:
  when invalid jdk is detected
  - strict
    - warning to console
    - strict mode -  dialogue popup, with "run: yes/no"
  - nostrict,
   - only warning to console

Corrct jdk detected
  - jsut info to console.



hmm? Unles you reply somehow soon, I will update patch with all above


For future:
  - I think ITW needs some mechanism to rember decision on those swarms of dialogues
  - some big work here will be needed and I started to work on prototype


J.