FYI: rework previous fix

[Aleksey Shipilev]

On 09/27/2017 05:45 PM, Roland Westrelin wrote:
>> This seems to revert our differences around has_only_data_users()
>> against upstream in favor of this one-liner, right?
> 
> Not quite. This below is new:
> 
> -        int alias = phase->C->get_alias_index(adr_type());
> -        Node* mem = find_mem_phi(n_loop_head, alias, phase->C);
> -        if (mem == NULL) {
> -          mem = in(Memory);
> -        }
> -

Okay then. The actual fix is still within the Shenandoah-specific code, and so we are good.

>> Make sure you have descriptive commit message!
> 
> Won't history all be lost when things are upstreamed so we should rather
> put comments?

It would be lost, right. But I'm mostly concerned with tracking the changes flowing between our
repositories at this point. Having a commit message "Revert unsuccessful fix for WB hoisting,
replace it with <...>" makes it stand out.

Thanks,
-Aleksey