[modules-dev] [Fwd: [Fwd: Review request: 6563535, URLRepository.install]]

[Dave Bristor]

Stanley M. Ho wrote:
> 
> Dave Bristor wrote:
>> Because ModuleArchiveInfo.equals() is asymmetric with respect to 
>> ModuleArchiveInfo.hashCode().  The former is based on equality of 
>> references and the latter on equality of hashcodes of a a subset of 
>> the instance's fields.  Here's the use case that necessitates this: 
>> From JRepo's uninstall command, 
>> URLRepository.uninstall(ModuleArchiveInfo) is given an instance that 
>> is created based on information from the command line, therefore the 
>> ModuleArchiveInfo instance is not the one returned by an earlier 
>> URLRepository.install().
> 
> The way I expected this to work is that Repository.list() will be used 
> to obtain a list of ModuleArchiveInfo, then we'll find the one we want 
> and pass it to uninstall(). I don't think the repository implementation 
> should attempt to recognize any arbitrary ModuleArchiveInfo that is 
> passed in. In other words, it should only recognize the ones that it 
> created.

OK, that makes sense, so I updated the code to take that into account.  In 
fact, I looked into JRepo's use of uninstall, and seems that will work as per 
your expectations above :-)

>> Now I understand.  Fixed, hopefully better, let me know what you 
>> think; I updated the webrev:
>>     http://analemma.sfbay.sun.com/java/jdk/ws/libs/rev/6563535-b/
> 
> Better. Note that the code uses writeRepositoryMetadata to write the 
> metadata. However, writeRepositoryMetadata() assumes that the new 
> ModuleArchiveInfo is in contents, and it simply iterates the contents 
> and writes the metadata out. Now we no longer update the contents 
> prematurely, so writeRepositoryMetadata() will need to call 
> writeModule() on the "mai" that is passed in.

Now there's a better-than-just-informational use for the mai that's passed in 
:-).  To make it work for both install and uninstall cases, I added a boolean 
to indicate if the given mai should be writeModule'd.

Same webrev updated again; tested on Solaris & Windows.

Thanks,
	Dave

> 
> - Stanley