Request for review (S) 7009828: Fix for 6938627 breaks visualvm monitoring when -Djava.io.tmpdir is defined

[Alan Bateman]

David Holmes wrote:
> :
> Understood but we're still left in a situation where it is unclear 
> what "components" will put what files where and whether it can, or 
> should be definable by the user.
>
> The CRs that motivated 6938627 are effectively all re-opened.
>
Yep, these issues need to be re-examined. As I recall, the motivation 
for the relatively recent change was when porting to platforms that 
didn't use /tmp and for that case it's probably best to deal with it as 
a porting time issue. As regards where the files go, this is currently 
something private to the specific mechanism, meaning the VM and tool/API 
side of jvmstat have to agree, the VM and tool/API side to the attach 
mechanism have a different agreement. It's unfortunate but difficult to 
change to a different means to locate and/or register endpoints, at 
least not without providing some backward compatibility and work with 
the HSX model.

-Alan.