[rfc][icedtea-web] Allow cache directory to be changed during runtime

[Jie Kang]

----- Original Message -----
> On 01/15/2015 07:16 PM, Andrew Azores wrote:
> > On 01/05/2015 02:49 PM, Jie Kang wrote:
> >> Hello,
> >>
> >> This patch consolidates the cacheDir instance to a single place and allows
> >> us to change it during
> >> run-time. It also modifies a unit test to use this new feature. This
> >> feature is aimed mainly at
> >> unit tests that need the cache, but do not want to modify the
> >> icedtea-web's user cache.
> >>
> >> Thoughts? Okay to push?
> 
> I'm wondering why it is needed at all.
> 
> JNLPRuntime.getConfiguration().getProperty(DeploymentConfiguration.KEY_USER_CACHE_DIR);
> is the place
> where you can set the dir.
> 
> 
> For unittest you may force refelction to do the job. if the chnage of the
> property isnot enough.
> 
> 
> What I would rather see is, to get rid of all the static stuff in
> CacheLRUWrapper and have
> possibility to create in instance.
> This instance will be also much more testable. But maybe it is overkill.

Hello,

Yes this is mainly to address that for unit tests, JNLPRuntime.[...] doesn't work since CacheLRUWrapper and CacheUtil both use static variables that get initialized on first use of either class. I think getting rid of all the static stuff in LRUWrapper and making it instantiable is a great idea. I see this patch as a baby-step towards doing that.

I very strongly prefer NOT using reflection to set the cacheDir value in the unit tests.

Also, please see [1] for a longer explanation

[1] http://mail.openjdk.java.net/pipermail/distro-pkg-dev/2015-January/030470.html


Regards,


> 
> J.
> >>
> >>
> >> Regards,
> >>
> >
> > I think this looks generally ok. Maybe as a separate enhancement changeset,
> > but since you're working
> > in here anyway, could you look at doing some refactoring to store the cache
> > directory as something
> > better than a String? As a File, or using Java 7's nio2 stuff, a Path
> > maybe?
> 
> On one side - yes, on the other - why? What operations are mainly done on the
> "string" string ops or
> file ops - I have not checked code, but seems like concate is most often :)
> 
> But yes - it is directory, so file or path is probably better then string.
> Please, do this refactoring as separate changset anyway.
> 
> J.
> 

-- 

Jie Kang

OpenJDK Team - Software Engineering Intern