Review request for JDK-6760902: inconsistent behavior in bootstrap class loader for classes and resources

[Mandy Chung]

On 1/16/2014 7:39 AM, Paul Sandoz wrote:
> On Jan 16, 2014, at 2:27 AM, Mandy Chung <mandy.chung at oracle.com> wrote:
>
>> There is an inconsistency in searching classes vs resources if bootclasspath contains an empty path.  Empty path on bootclasspath is skipped by the bootstrap class loader when searching classes while it defaults to current working directory when searching resources as the application class loader.
> Just curious, where is the code that skips empty paths in the bootclasspath when loading classes? Is that in the VM?

Yes it's in the VM which implements the class loading.   I don't know 
the history and I guess splitting the implementation of class loading vs 
resource lookup in the VM and jdk could be due to no actual ClassLoader 
implementation (null instead).

>>   This fixes sun.misc.Launcher to skip empty path when constructing the paths from bootclasspath for resource lookup.
>>
>> Webrev at:
>> http://cr.openjdk.java.net/~mchung/jdk9/webrevs/6760902/webrev.00/
>>
> Perhaps it is better to merge GetResource2.sh into GetResource.sh.

I was considering if merging them would be better and separating it 
makes it clear for what it tests for.  On the other hand, I don't like 
duplication and I probably can merge them and clean

> I am so tempted to change this too :-)
>
>   423                 if (pos - lastPos > 0) {

Me too and time to do it.  Will post another webrev.

Mandy
>
> Paul.
>
>
>> There is some incompatibility risk that may impact existing code depending on this behavior to search resources from the current working directory if empty path (rather than explicit) is set.  I think most application using bootclasspath is to add their paths to load their classes and likely expect the resources are looked in the consistent way (i.e. skips the empty path).  So I expect the incompatibility risk is low.
>>
>> thanks
>> Mandy