> On June 12, 2013, 11:43 p.m., Vinod Kone wrote:
> > 3rdparty/libprocess/3rdparty/stout/include/stout/os/killtree.hpp, lines 
> > 110-113
> > <https://reviews.apache.org/r/11849/diff/1/?file=304306#file304306line110>
> >
> >     You only add one process to 'stopped'/'visited' right in one iteration 
> > right?
> >     
> >     Why are you repeatedly sending SIGSTOP to the same process?
> >     
> >     Also, I highly recommend adding a verbose option to this function. It 
> > has proved invaluable to us when using/debugging the erstwhile killtree 
> > script.

This only sends SIGSTOP to each process once.

If someone calls killtree with SIGSTOP then killtree behaves strangely. Perhaps 
I should guard against that, thoughts?


> On June 12, 2013, 11:43 p.m., Vinod Kone wrote:
> > 3rdparty/libprocess/3rdparty/stout/include/stout/os/killtree.hpp, line 79
> > <https://reviews.apache.org/r/11849/diff/1/?file=304306#file304306line79>
> >
> >     I think it makes more sense to call 'stopped' set as 'visited' per 
> > typical graph algo terminology. I was a bit confused because i was equating 
> > stopped with processing getting a SIGSTOP.

The nice thing about stopped is it clearly captures when each process has been 
sent SIGSTOP. I initially went with a generic BFS implementation in terms of 
terminology and insertion into 'visited' but it's not as clear I found.


- Ben


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/11849/#review21824
-----------------------------------------------------------


On June 12, 2013, 7:49 p.m., Ben Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/11849/
> -----------------------------------------------------------
> 
> (Updated June 12, 2013, 7:49 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman and Vinod Kone.
> 
> 
> Description
> -------
> 
> For how killtree will be used, see the killtree.sh script in mesos.
> 
> This version is actually more robust than killtree.sh, as it first stops the 
> entire tree, _then_ issues the signal.
> Currently, killtree.sh stops, signals, and continues each process in 
> isolation.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/3rdparty/stout/include/stout/os.hpp 
> 1b3fb47d7567b5467fef2a2bb15d5c4a2ea42aa5 
>   3rdparty/libprocess/3rdparty/stout/include/stout/os/killtree.hpp 
> PRE-CREATION 
>   3rdparty/libprocess/3rdparty/stout/tests/os_tests.cpp 
> 047778d05ebbbefd85e4a163dbb6ab8445edfb7f 
> 
> Diff: https://reviews.apache.org/r/11849/diff/
> 
> 
> Testing
> -------
> 
> Added a test, however this test can be _significantly_ cleaned up were we to 
> create an abstraction for generating process trees!
> 
> 
> Thanks,
> 
> Ben Mahler
> 
>

Reply via email to