[ 
https://issues.apache.org/jira/browse/MESOS-5886?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16395580#comment-16395580
 ] 

Andrei Budnik commented on MESOS-5886:
--------------------------------------

If a class contains methods with the same signature, we can add one extra 
argument of a unique type for each of these methods. We can introduce a special 
type `Tag` for extra arguments:
{code:java}
struct None {};

template <int val>
struct Tag { Tag(const None&) {} };

class Example : public Process<Example> {
public:
  void m1(int, Tag<1> tag = None()) {}
  void m2(int, Tag<2> tag = None()) {}
  void m3(int, Tag<3> tag = None()) {}
};{code}
It looks like current implementation of 
[`dispatch()`|https://github.com/apache/mesos/blob/8adb5fcb1f6c451bc9ad7ecdc6e39bc170fdcd65/3rdparty/libprocess/include/process/dispatch.hpp#L209-L256]
 supports arguments with a default value, so a user of `dispatch()` doesn't 
need to care about the extra tag argument.
Note, that this approach fixes the issue only for non-virtual methods. For more 
details, see [On fixing 
FUTURE_DISPATCH|https://docs.google.com/document/d/1d4nYfIWuTGtvHObolyzCwBttU75gxbKSvyISjHQ76Nw]
 document.

 

> FUTURE_DISPATCH may react on irrelevant dispatch.
> -------------------------------------------------
>
>                 Key: MESOS-5886
>                 URL: https://issues.apache.org/jira/browse/MESOS-5886
>             Project: Mesos
>          Issue Type: Bug
>    Affects Versions: 1.1.2, 1.2.1, 1.3.0, 1.4.0
>            Reporter: Alexander Rukletsov
>            Priority: Major
>              Labels: mesosphere, tech-debt, tech-debt-test
>
> [{{FUTURE_DISPATCH}}|https://github.com/apache/mesos/blob/e8ebbe5fe4189ef7ab046da2276a6abee41deeb2/3rdparty/libprocess/include/process/gmock.hpp#L50]
>  uses 
> [{{DispatchMatcher}}|https://github.com/apache/mesos/blob/e8ebbe5fe4189ef7ab046da2276a6abee41deeb2/3rdparty/libprocess/include/process/gmock.hpp#L350]
>  to figure out whether a processed {{DispatchEvent}} is the same the user is 
> waiting for. However, comparing {{std::type_info}} of function pointers is 
> not enough: different class methods with same signatures will be matched. 
> Here is the test that proves this:
> {noformat}
> class DispatchProcess : public Process<DispatchProcess>
> {
> public:
>   MOCK_METHOD0(func0, void());
>   MOCK_METHOD1(func1, bool(bool));
>   MOCK_METHOD1(func1_same_but_different, bool(bool));
>   MOCK_METHOD1(func2, Future<bool>(bool));
>   MOCK_METHOD1(func3, int(int));
>   MOCK_METHOD2(func4, Future<bool>(bool, int));
> };
> {noformat}
> {noformat}
> TEST(ProcessTest, DispatchMatch)
> {
>   DispatchProcess process;
>   PID<DispatchProcess> pid = spawn(&process);
>   Future<Nothing> future = FUTURE_DISPATCH(
>       pid,
>       &DispatchProcess::func1_same_but_different);
>   EXPECT_CALL(process, func1(_))
>     .WillOnce(ReturnArg<0>());
>   dispatch(pid, &DispatchProcess::func1, true);
>   AWAIT_READY(future);
>   terminate(pid);
>   wait(pid);
> }
> {noformat}
> The test passes:
> {noformat}
> [ RUN      ] ProcessTest.DispatchMatch
> [       OK ] ProcessTest.DispatchMatch (1 ms)
> {noformat}
> This change was introduced in https://reviews.apache.org/r/28052/.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to