> On Nov. 9, 2016, 3:44 a.m., Benjamin Mahler wrote: > > Unrelated to this particular change, but can you write up a quick test > > (doesn't need to be committed) to ensure that we don't have infinitely > > growing future chains here, as benh mentioned offline? I never tested this > > for http::Connection, but I suspect we may have a memory growth problem if > > we open a connection and loop sending requests on it (try with both pipe > > and body).
Sure, as discussed offline, would look up the memory usage of a example test framework. Many of them use the `Connection` abstraction under the hood. > On Nov. 9, 2016, 3:44 a.m., Benjamin Mahler wrote: > > 3rdparty/libprocess/src/http.cpp, lines 898-900 > > <https://reviews.apache.org/r/53489/diff/1/?file=1554498#file1554498line898> > > > > Do you want to clear the content length header if they are using a pipe? We already check that content-length is not set for piped requests and fail the request in `send()`. I would add a explicit invariant check. - Anand ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/53489/#review155395 ----------------------------------------------------------- On Nov. 4, 2016, 5:55 p.m., Anand Mazumdar wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/53489/ > ----------------------------------------------------------- > > (Updated Nov. 4, 2016, 5:55 p.m.) > > > Review request for mesos and Benjamin Mahler. > > > Bugs: MESOS-6466 > https://issues.apache.org/jira/browse/MESOS-6466 > > > Repository: mesos > > > Description > ------- > > This required modifications to the `encode()` method to return > a `Pipe::Reader` instead of the request body. The `send()` then > reads from this pipe to send the request via the socket. > > > Diffs > ----- > > 3rdparty/libprocess/src/http.cpp b5ddac9f1970358f8232a481431507242052d294 > > Diff: https://reviews.apache.org/r/53489/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Anand Mazumdar > >
