> On May 15, 2018, 8:28 p.m., Andrew Schwartzmeyer wrote:
> > src/slave/containerizer/mesos/launch.cpp
> > Lines 1110-1113 (patched)
> > <https://reviews.apache.org/r/67137/diff/1/?file=2023058#file2023058line1110>
> >
> >     Why not just `O_CLOEXEC` all the handles opened on Linux (so that you 
> > can avoid all this kind of hacky logic here)? (And fix the third-party libs 
> > to do the same.)
> 
> Benjamin Bannier wrote:
>     Third-party libraries might be dynamically loaded from the system in 
> unbundled configurations; we should support unbundled libraries. This makes 
> patching all dependencies unrealistic. A solution to address such 
> configuration would be to `LD_PRELOAD` an `::open` which we control and which 
> always sets `O_CLOEXEC`. The approach here tidies up what we expose to 
> external code.

Dropping this. Please reopen if I am missing something.


- Benjamin


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/67137/#review203154
-----------------------------------------------------------


On May 18, 2018, 2:36 p.m., Benjamin Bannier wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/67137/
> -----------------------------------------------------------
> 
> (Updated May 18, 2018, 2:36 p.m.)
> 
> 
> Review request for mesos, Gilbert Song, Jie Yu, and James Peach.
> 
> 
> Bugs: MESOS-8917
>     https://issues.apache.org/jira/browse/MESOS-8917
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Avoided leaking file descriptors in Mesos containerizer.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/mesos/launch.cpp 
> f25d90651ef32495c9161c3eaed8a327d1b2b926 
> 
> 
> Diff: https://reviews.apache.org/r/67137/diff/4/
> 
> 
> Testing
> -------
> 
> `make check`
> 
> I still need to confirm this patch in CI on a wider range of scenarios.
> 
> 
> Thanks,
> 
> Benjamin Bannier
> 
>

Reply via email to