Re: Review Request 60598: Replaced `abort()` with `_exit()` in `ChildHook::SUPERVISOR`.

[Andrei Budnik <ab...@mesosphere.com>]

> On July 5, 2017, 4:29 p.m., Benjamin Bannier wrote:
> > 3rdparty/libprocess/src/subprocess.cpp
> > Lines 176 (patched)
> > <https://reviews.apache.org/r/60598/diff/2/?file=1768981#file1768981line176>
> >
> >     Any reason we cannot do a proper `exit` here?
> >     
> >     Here and below.

1. `exit()` isn't async-signal-safe function.
2. Watchdog (i.e. parent) process should retransmit any failures from a child process, hence watchdog exits with `EXIT_FAILURE` in case of a child failure.


- Andrei


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/60598/#review179643
-----------------------------------------------------------


On July 3, 2017, 2:58 p.m., Andrei Budnik wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/60598/
> -----------------------------------------------------------
> 
> (Updated July 3, 2017, 2:58 p.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov, Benjamin Bannier, Benjamin Mahler, and James Peach.
> 
> 
> Bugs: MESOS-7160
>     https://issues.apache.org/jira/browse/MESOS-7160
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Previously, `abort()` was called in `SUPERVISOR` hook when child
> process exited with an error code, or `waitpid()` failed, or parent
> process exited. All these cases shouldn't lead to abnormal program
> termination with coredumps.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/src/subprocess.cpp 0f1532b294d6d6b1e017468cfde47362f3faa84d 
> 
> 
> Diff: https://reviews.apache.org/r/60598/diff/3/
> 
> 
> Testing
> -------
> 
> make check (mac os x, fedora 25)
> 
> 
> Thanks,
> 
> Andrei Budnik
> 
>