You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Neil Conway <ne...@gmail.com> on 2016/04/11 03:41:11 UTC

Review Request 45996: Fixed memory leak of `gc` in `finalize()` in libprocess.

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45996/
-----------------------------------------------------------

Review request for mesos and Joris Van Remoortere.


Bugs: MESOS-5144
    https://issues.apache.org/jira/browse/MESOS-5144


Repository: mesos


Description
-------

Fixed memory leak of `gc` in `finalize()` in libprocess.


Diffs
-----

  3rdparty/libprocess/include/process/gc.hpp 799468ebe49f2a49d325f40ffd8acea727abf74c 
  3rdparty/libprocess/src/process.cpp 5e9dcfdc52f3a8223bc43af149b8e1f5dbdf5b0a 

Diff: https://reviews.apache.org/r/45996/diff/


Testing
-------


Thanks,

Neil Conway

Re: Review Request 45996: Fixed memory leak of `gc` in `finalize()` in libprocess.

Posted by Neil Conway <ne...@gmail.com>.

> On April 14, 2016, 7:10 p.m., Joseph Wu wrote:
> > 3rdparty/libprocess/src/process.cpp, lines 489-490
> > <https://reviews.apache.org/r/45996/diff/1/?file=1338532#file1338532line489>
> >
> >     This is now effectively managed by the `ProcessManager`, so you could bring it into the `ProcessManager` 's scope (which would still be global).
> 
> Neil Conway wrote:
>     This makes sense (and I would love to eliminate all the global state in libprocess), but AFAICS it won't be easy to implement: if we make `gc` a field of `ProcessManager`, we'd ideally like to initialize and spawn the GC process in `ProcessManager`'s constructor. But that isn't possible, because we can't safely call `spawn` at the time we are constructing the ProcessManager. We could still move `gc` to be a field of `ProcessManager` but spawn it separately, but I'm not sure if that is actually a net improvement at that point.
> 
> Joseph Wu wrote:
>     It should be safe to call `ProcessManager::spawn` rather than `process::spawn` for `gc`.  (Unless there's some dependency on the socket `__s__` for new processes...)

Sure, but it still seems a bit fragile to me. I'd prefer to take on the removal of all the libprocess global state as a separate project.


- Neil


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45996/#review128956
-----------------------------------------------------------


On April 11, 2016, 1:41 a.m., Neil Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45996/
> -----------------------------------------------------------
> 
> (Updated April 11, 2016, 1:41 a.m.)
> 
> 
> Review request for mesos and Joris Van Remoortere.
> 
> 
> Bugs: MESOS-5144
>     https://issues.apache.org/jira/browse/MESOS-5144
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Fixed memory leak of `gc` in `finalize()` in libprocess.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/gc.hpp 799468ebe49f2a49d325f40ffd8acea727abf74c 
>   3rdparty/libprocess/src/process.cpp 5e9dcfdc52f3a8223bc43af149b8e1f5dbdf5b0a 
> 
> Diff: https://reviews.apache.org/r/45996/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Neil Conway
> 
>

Re: Review Request 45996: Fixed memory leak of `gc` in `finalize()` in libprocess.

Posted by Joseph Wu <jo...@mesosphere.io>.

> On April 14, 2016, 12:10 p.m., Joseph Wu wrote:
> > 3rdparty/libprocess/src/process.cpp, lines 489-490
> > <https://reviews.apache.org/r/45996/diff/1/?file=1338532#file1338532line489>
> >
> >     This is now effectively managed by the `ProcessManager`, so you could bring it into the `ProcessManager` 's scope (which would still be global).
> 
> Neil Conway wrote:
>     This makes sense (and I would love to eliminate all the global state in libprocess), but AFAICS it won't be easy to implement: if we make `gc` a field of `ProcessManager`, we'd ideally like to initialize and spawn the GC process in `ProcessManager`'s constructor. But that isn't possible, because we can't safely call `spawn` at the time we are constructing the ProcessManager. We could still move `gc` to be a field of `ProcessManager` but spawn it separately, but I'm not sure if that is actually a net improvement at that point.

It should be safe to call `ProcessManager::spawn` rather than `process::spawn` for `gc`.  (Unless there's some dependency on the socket `__s__` for new processes...)


- Joseph


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45996/#review128956
-----------------------------------------------------------


On April 10, 2016, 6:41 p.m., Neil Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45996/
> -----------------------------------------------------------
> 
> (Updated April 10, 2016, 6:41 p.m.)
> 
> 
> Review request for mesos and Joris Van Remoortere.
> 
> 
> Bugs: MESOS-5144
>     https://issues.apache.org/jira/browse/MESOS-5144
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Fixed memory leak of `gc` in `finalize()` in libprocess.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/gc.hpp 799468ebe49f2a49d325f40ffd8acea727abf74c 
>   3rdparty/libprocess/src/process.cpp 5e9dcfdc52f3a8223bc43af149b8e1f5dbdf5b0a 
> 
> Diff: https://reviews.apache.org/r/45996/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Neil Conway
> 
>

Re: Review Request 45996: Fixed memory leak of `gc` in `finalize()` in libprocess.

Posted by Neil Conway <ne...@gmail.com>.

> On April 14, 2016, 7:10 p.m., Joseph Wu wrote:
> > 3rdparty/libprocess/src/process.cpp, lines 489-490
> > <https://reviews.apache.org/r/45996/diff/1/?file=1338532#file1338532line489>
> >
> >     This is now effectively managed by the `ProcessManager`, so you could bring it into the `ProcessManager` 's scope (which would still be global).

This makes sense (and I would love to eliminate all the global state in libprocess), but AFAICS it won't be easy to implement: if we make `gc` a field of `ProcessManager`, we'd ideally like to initialize and spawn the GC process in `ProcessManager`'s constructor. But that isn't possible, because we can't safely call `spawn` at the time we are constructing the ProcessManager. We could still move `gc` to be a field of `ProcessManager` but spawn it separately, but I'm not sure if that is actually a net improvement at that point.


- Neil


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45996/#review128956
-----------------------------------------------------------


On April 11, 2016, 1:41 a.m., Neil Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45996/
> -----------------------------------------------------------
> 
> (Updated April 11, 2016, 1:41 a.m.)
> 
> 
> Review request for mesos and Joris Van Remoortere.
> 
> 
> Bugs: MESOS-5144
>     https://issues.apache.org/jira/browse/MESOS-5144
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Fixed memory leak of `gc` in `finalize()` in libprocess.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/gc.hpp 799468ebe49f2a49d325f40ffd8acea727abf74c 
>   3rdparty/libprocess/src/process.cpp 5e9dcfdc52f3a8223bc43af149b8e1f5dbdf5b0a 
> 
> Diff: https://reviews.apache.org/r/45996/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Neil Conway
> 
>

Re: Review Request 45996: Fixed memory leak of `gc` in `finalize()` in libprocess.

Posted by Joseph Wu <jo...@mesosphere.io>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45996/#review128956
-----------------------------------------------------------


Fix it, then Ship it!





3rdparty/libprocess/src/process.cpp (lines 489 - 490)
<https://reviews.apache.org/r/45996/#comment192409>

    This is now effectively managed by the `ProcessManager`, so you could bring it into the `ProcessManager` 's scope (which would still be global).


- Joseph Wu


On April 10, 2016, 6:41 p.m., Neil Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45996/
> -----------------------------------------------------------
> 
> (Updated April 10, 2016, 6:41 p.m.)
> 
> 
> Review request for mesos and Joris Van Remoortere.
> 
> 
> Bugs: MESOS-5144
>     https://issues.apache.org/jira/browse/MESOS-5144
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Fixed memory leak of `gc` in `finalize()` in libprocess.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/gc.hpp 799468ebe49f2a49d325f40ffd8acea727abf74c 
>   3rdparty/libprocess/src/process.cpp 5e9dcfdc52f3a8223bc43af149b8e1f5dbdf5b0a 
> 
> Diff: https://reviews.apache.org/r/45996/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Neil Conway
> 
>