I'm sorry I didn't notice pair "acquireShared" and "releaseShared". Okay,
well, there are no races. But we still can use "getState()" in "assert"
instead "resFlag", that will allow us use boolean type. But if you think it
will not make code cleaner, then I will close issue.
2017-03-01 14:55 GMT+03
First of all I want to replace the "byte resFlag" on the "bool haveResult"
only for aesthetic reasons. I think that use a byte as a boolean flag looks
like C in 1983. And constructions like
return haveResult ? (R)res : null;
more readable unlike
return resFlag == RES ? (R)res : null;
because yo
Alexander,
Can you please clarify this?
>I suppose "volatile" is not need here, because right now if
>GridFutureAdapter#onDone()
called in another thread than GridFutureAdapter#get0, it will produce
AssertionError.
1) 99.999% of time GridFutureAdapter#onDone() is called in some other
thread tha
Sorry, I mean I want to do fast work on this issue (not make code faster).
Your code is strange. You can see my view in my local temp PR:
https://github.com/SharplEr/ignite/pull/4/files
This is what I'm meaning.
I suppose "volatile" is not need here, because right now if
GridFutureAdapter#onDone
Alexey, Vladislav, are you agree with me, or not? I want to do it faster
and move on.
2017-02-17 17:36 GMT+03:00 Александр Меньшиков :
> We can check "onDone" method was called with using getState() method. If
> getState()!=0 then resFlag!=0. Just look at code.
>
>
> 2017-02-17 17:12 GMT+03:00 А
We can check "onDone" method was called with using getState() method. If
getState()!=0 then resFlag!=0. Just look at code.
2017-02-17 17:12 GMT+03:00 Александр Меньшиков :
> Like I said, if "resFlag==0" (of course 0 came from initialization) means
> "you still haven't called the method onDone()"
Like I said, if "resFlag==0" (of course 0 came from initialization) means
"you still haven't called the method onDone()", better make it clear.
2017-02-17 14:48 GMT+03:00 Vladislav Pyatkov :
> Alexander,
>
> I think, the resFlag will be initiated as 0 (new GridFutureAdapter()), but
> 1 and 2 sta
Alexander,
I think, the resFlag will be initiated as 0 (new GridFutureAdapter()), but
1 and 2 states will be acquired on live.
On Fri, Feb 17, 2017 at 1:56 PM, Александр Меньшиков
wrote:
> Alexey,
>
> I see only one place where writes in resFlag:
>
> if (err != null) {
>
Alexey,
I see only one place where writes in resFlag:
if (err != null) {
resFlag = ERR;
this.res = err;
}
else {
resFlag = RES;
this.res = res;
}
And th
Alexander,
This change is not applicable for GridFutureAdapter because resFlag can
have 3 values there.
2017-02-16 19:58 GMT+03:00 Александр Меньшиков :
> Hello.
>
> I propose to do refactoring of classes "GridFinishedFuture" and
> "GridFutureAdapter". There is field "resFlag" which can equals "
Hello.
I propose to do refactoring of classes "GridFinishedFuture" and
"GridFutureAdapter". There is field "resFlag" which can equals "ERR = 1" or
"RES = 2". So I can replace it to one "bool haveResult" field.
If there are no objections, I'm ready to proceed. If you find more such
classes, please
Alexander Menshikov created IGNITE-4713:
---
Summary: refactoring of GridFinishedFuture and GridFutureAdapter
Key: IGNITE-4713
URL: https://issues.apache.org/jira/browse/IGNITE-4713
Project: Ignite
12 matches
Mail list logo
