> On Feb. 26, 2015, 6:14 p.m., Ben Mahler wrote: > > This is pretty inconsistent with the rest of our code. Why can't you add > > the missing deletes? > > Dominic Hamon wrote: > we can. but why not start getting rid of the archaic practice of using > raw pointers? > > Timothy Chen wrote: > I agree having less prone to errors is more desired.
>From the title, it looks like this would have changed all buffers in os.hpp, >rather than introducing inconsistency in the process of fixing the two leaks >(detected by coverity). More importantly, is '`vector`' an intuitive way to capture a fixed size buffer on the heap? I'm all for improving this, but we need to think more carefully about how to represent buffers, do we want to introduce some kind of '`Buffer`' abstraction if benefical to readability (e.g. implicitly castable to the pointer type, etc), or consider `Owned<>`, `unique_ptr`, `scoped_ptr`, etc? Let's do these kind of changes independently. - Ben ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/31489/#review74305 ----------------------------------------------------------- On Feb. 26, 2015, 6:14 p.m., Dominic Hamon wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/31489/ > ----------------------------------------------------------- > > (Updated Feb. 26, 2015, 6:14 p.m.) > > > Review request for mesos, Joerg Schad and Till Toenshoff. > > > Bugs: MESOS-2412 > https://issues.apache.org/jira/browse/MESOS-2412 > > > Repository: mesos > > > Description > ------- > > Fix use of buffers in OS to remove leaks. > > > Diffs > ----- > > 3rdparty/libprocess/3rdparty/stout/include/stout/os.hpp > 8a4fda97ee29c185471a69f60803efc193cbe922 > > Diff: https://reviews.apache.org/r/31489/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Dominic Hamon > >
