> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 858
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line858>
> >
> >     This is great, but can we think of any other abstractions this can use 
> > to make this stuff even simpler? In particular, what about a 
> > process::io::read which handles spurious events? See this: 
> > https://reviews.apache.org/r/5567

Done.


> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 870
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line870>
> >
> >     virtual

Done.


> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 872
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line872>
> >
> >     virtual

Done.


> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 933
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line933>
> >
> >     Is this supposed to account for spurious events? Or couldn't "none" 
> > here mean EOF?

Done.


> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 383
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line383>
> >
> >     This is so similar to just utils::os::read that we should create 
> > another utils version that does the conversion for you (probably via 
> > std::stringstream or boost::lexical_cast). The big difference here is 
> > returning 'none' for EAGAIN, but see my comment below.

Done.


> On June 25, 2012, 8:01 p.m., Benjamin Hindman wrote:
> > src/linux/cgroups.cpp, line 328
> > <https://reviews.apache.org/r/5395/diff/5/?file=112400#file112400line328>
> >
> >     Older versions of glibc don't have eventfd. We need to wrap it if 
> > necessary.

Maybe we should add a check in the autotool to check the glibc version. But 
anyway, cgroups is not available in the kernel if eventfd is not.

Therefore, I think there are two possibilities if eventfd is not available:

1) if you try to compile on that platform, compilation will not succeed
2) if you compile on other machine and try to run the binary on a machine 
without eventfd, the linking will fail


- Jie


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/5395/#review8426
-----------------------------------------------------------


On June 26, 2012, 6:21 p.m., Jie Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/5395/
> -----------------------------------------------------------
> 
> (Updated June 26, 2012, 6:21 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman and Vinod Kone.
> 
> 
> Description
> -------
> 
> In cgroups, sometimes you want to be notified if some certain events happen. 
> For example, users may want to be notified when an out-of-memory event 
> happens.
> 
> This patch introduce a function called "listenEvent" which returns an 
> instance of Future so that you can check the status of this future to see 
> whether any event happens.
> 
> This interface is general to all the events in cgroups.
> 
> 
> Diffs
> -----
> 
>   src/linux/cgroups.hpp PRE-CREATION 
>   src/linux/cgroups.cpp PRE-CREATION 
>   src/tests/cgroups_tests.cpp PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/5395/diff/
> 
> 
> Testing
> -------
> 
> On Linux machines, make check.
> 
> 
> Thanks,
> 
> Jie Yu
> 
>

Reply via email to