On Thu, May 9, 2013 at 5:38 AM, Alan Bateman <alan.bate...@oracle.com>wrote:
> On 08/05/2013 23:05, Martin Buchholz wrote:
>
>> Alan (et al): Ping.
>>
>> I've looked through the webrev.
>
> The scary part is subsequenceRaw where the offsets including the position.
> I don't see anything obviously wrong and the tests should catch any issues.
> I don't see any issue conditionally generating the isDirect/order methods
> although it should been harmless.
>
Yes, it's scary, but it's much scarier to me carrying around relative
positions than absolute because position() can change at any time if there
is a data race. So after bounds checking we must work with absolute
positions, or at least must when working with Unsafe.
---
Existing code is a little schizoid about checking for the possibility of
limit() < position(), as can happen with a data race. Sometimes checked;
sometimes not. Sometimes assert()ed. Sometimes adjusted. Probably the
right thing is to check and throw ConcurrentModificationException if
detected. But not in this changeset.
int pos = this.position();
int lim = this.limit();
assert (pos <= lim);
int rem = (pos <= lim ? lim - pos : 0);
