> On May 17, 2018, 11:21 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/stout/include/stout/internal/windows/inherit.hpp
> > Lines 34 (patched)
> > <https://reviews.apache.org/r/67201/diff/1/?file=2024989#file2024989line34>
> >
> >     What do the second and third arguments represent?

Second argument is the number of attributes you want to update. The third one 
is reserved(windows... sigh)


> On May 17, 2018, 11:21 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/stout/include/stout/internal/windows/inherit.hpp
> > Lines 41-43 (patched)
> > <https://reviews.apache.org/r/67201/diff/1/?file=2024989#file2024989line41>
> >
> >     We should probably use a `std::unique_ptr<PROC_THREAD_ATTRIBUTE_LIST>` 
> > here for RAII.

Defined an std::vector since _PROC_THREAD_ATTRIBUTE_LIST is not defined in a 
public header


> On May 17, 2018, 11:21 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/stout/include/stout/internal/windows/inherit.hpp
> > Lines 50 (patched)
> > <https://reviews.apache.org/r/67201/diff/1/?file=2024989#file2024989line50>
> >
> >     RAII takes care of this ;)
> >     
> >     I noticed below we also called `DeleteProcThreadAttributeList`. Under 
> > what conditions do we `DeleteProcThreadAttributeList` and `free`, and when 
> > do we just `free`?

DeleteProcThreadAttributeList is used to release memory when 
proc_thread_attribute_list is initialized successfully. You have to call free 
to deallocate it(due to the malloc)


> On May 17, 2018, 11:21 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/stout/include/stout/os/windows/shell.hpp
> > Line 269 (original), 270 (patched)
> > <https://reviews.apache.org/r/67201/diff/1/?file=2024990#file2024990line270>
> >
> >     Is there a `W` version of this, or is that redundant...? I don't know.

If UNICODE is defined(and it is), the function names are automatically suffixed 
wth W. I changed to W version to be consistent for now.


> On May 17, 2018, 11:21 p.m., Andrew Schwartzmeyer wrote:
> > 3rdparty/stout/include/stout/os/windows/shell.hpp
> > Lines 279-291 (original), 283-306 (patched)
> > <https://reviews.apache.org/r/67201/diff/1/?file=2024990#file2024990line283>
> >
> >     (Old comment: Since the size is `constexpr` (it's always 3), I don't 
> > think we need to malloc on the heap. We can just use `std::array<HANDLE, 
> > 3>`.)
> >     
> >     Actually, I don't think you even need `handle_array` at all. `pipes` is 
> > already a `std::array<int_fd, 3>` (and `int_fd` in this case is `HANDLE`). 
> > You can use the 
> > [data()](http://en.cppreference.com/w/cpp/container/array/data) member of 
> > `std::array` to get access to the underlying `HANDLE[]`.
> 
> Andrew Schwartzmeyer wrote:
>     Oops, I was rushing. `int_fd` is castable to `HANDLE` but is an 
> abstracted data type that has another field. So you'd need 
> `std::vector<HANDLE> handles(pipes->cbegin(), pipes->cend())` and then pass 
> `handles.data()`. But still, allocated automatically with no need to rfree :)

Thanks.


- Radhika


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/67201/#review203382
-----------------------------------------------------------


On May 17, 2018, 10:18 p.m., Radhika Jandhyala wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/67201/
> -----------------------------------------------------------
> 
> (Updated May 17, 2018, 10:18 p.m.)
> 
> 
> Review request for mesos and Andrew Schwartzmeyer.
> 
> 
> Bugs: MESOS-8926
>     https://issues.apache.org/jira/browse/MESOS-8926
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Whitelist handles that can be inherited by a child process.
> 
> 
> Diffs
> -----
> 
>   3rdparty/stout/include/stout/internal/windows/inherit.hpp 
> 7dbde820e775cbaeb8db4bc4559ab432903e75ea 
>   3rdparty/stout/include/stout/os/windows/shell.hpp 
> 8da612af2888ff4d4d458ea5b16cdb08779b6f4c 
> 
> 
> Diff: https://reviews.apache.org/r/67201/diff/1/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Radhika Jandhyala
> 
>

Reply via email to