Resolver and inefficiencies

[Paul Sandoz]

On May 3, 2012, at 6:58 PM, Alan Bateman wrote:
> On 03/05/2012 16:13, Paul Sandoz wrote:
>> Hi,
>> 
>> This observation may already be known, but i thought i would throw this out here anyway...
>> 
> Yes, it's a known problem and the resolver needs work. There was an interesting thread between Jaroslav Tulach and Jeff Glick a few months ago on this topic.

OK, i will look through the archives.


> While the resolver is only run today at install time,

Even then it could potentially blow up javac's/jmod's stack :-)


> it will eventually need to be run at runtime, say with containers deploying applications at runtime and that usage will clearly need to be very efficient.
> 

Yes, it seems doable (beyond the obvious of caching the result of resolving certain stuff, like jdk.base).

Paul.