RFR: 8000617: It should be possible to allocate memory without the VM dying.

[Nils Loodin]

A summary of what has happened so far, since I believe the major points 
have been discussed.

John Rose was worried of the added usages of std::, but actually 
removing all of those existing (if appropriate!) would be a separate, 
bigger, change. This is only fixing an outage in a way that' coherent 
with what's already here.

Furthermore, David Holmes has worried about

1. Replacing os::malloc with AllocateHeap, to which me and Coleen has 
argued this intially is a symptom of above mentioned outage, and it will 
add to a better tracing of allocations.

2. That there's an Enum instead of a const bool, to which many have 
argued that this increases type safety, and disallows people 
accidentally sending the wrong boolean.

3. That the name AllocFailStrategy is too long. To which I've rather 
un-eloquently responded 'no it's not, nya nya'.
I think it's the best and most descriptive one I've seen anyway.

David has said that the change webrev.03 does not explicitly go against 
his vision of 'path of least resistance', which I take to mean that it's 
pretty much ok to him anyway.

Coleen and Keith has said that webrev.03 is ok, minus the unfortunate 
mistaken change from nothrow_constant to throw_constant.

Here's a webrev without the error:
http://cr.openjdk.java.net/~nloodin/8000617/webrev.06/

If I missed an important point, please let me know!

Regards,
Nils Loodin