> On Sept. 23, 2016, 1:31 a.m., Joris Van Remoortere wrote: > > Why this solution as opposed to ensuring we initialize at the beginning of > > processes like we do with `process::initialize()`? > > Benjamin Bannier wrote: > Fixing the code in this spot is slightly easier as `openssl::initialize` > would > need to be called from a number of sites. Having it being called as part > of > `process::initialize` would be a viable option, but would make testing > slightly > harder.
I filed https://issues.apache.org/jira/browse/MESOS-6328 to make openssl initialization eager. - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/52154/#review150098 ----------------------------------------------------------- On Sept. 26, 2016, 6:16 p.m., Benjamin Bannier wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/52154/ > ----------------------------------------------------------- > > (Updated Sept. 26, 2016, 6:16 p.m.) > > > Review request for mesos, Joris Van Remoortere and Till Toenshoff. > > > Bugs: MESOS-6216 > https://issues.apache.org/jira/browse/MESOS-6216 > > > Repository: mesos > > > Description > ------- > > This code modified the process environment in order to support upgrading > on the fly from old-style libprocess SSL variables `SSL_` to > `LIBPROCESS_SSL_`. Modifying the process environment at this point is > not safe as other actors might concurrently read out that same > environment. > > Instead avoid changing the process environment altogether since flags > can just as well be read from a map. > > > Diffs > ----- > > 3rdparty/libprocess/src/openssl.cpp > c09cdc89509e4e4ca4c8a0f4fb0a57156a3a6091 > > Diff: https://reviews.apache.org/r/52154/diff/ > > > Testing > ------- > > Tested on various platforms in internal CI. > > > Thanks, > > Benjamin Bannier > >
