[nanomsg] Re: [announce] nanomsg 0.8-beta released

  • From: Stefano <phd.st.p@xxxxxxxxx>
  • To: nanomsg@xxxxxxxxxxxxx
  • Date: Sun, 22 Nov 2015 13:48:24 +0000

On 20 November 2015 at 17:37, Garrett D'Amore <garrett@xxxxxxxxxx> wrote:

All,

I’ve just posted release 0.8 beta of nanomsg.

Its at the usual location: http://nanomsg.org/download.html (or rather the
github releases page: https://github.com/nanomsg/nanomsg/releases )

Thank you very much for your hard work!


This is a bug fix release, with a number of problematic bugs resolved. Most
notably, anyone trying to use the raw mode API with REQ/REP sockets (and
that includes nn_device type usage!) has probably seen either completely or
partially broken behavior in the past. I believe these issues are now
resolved.

I am indeed using nn_device() for a broker model (a single node gets
all the REQ/REP requests, dispatches them to the workers, and send
back the results to the clients).

What kind of issues could I have experienced?

Something that happened to me in the past is that if I mistakenly set
a NN_REQ_RESEND_IVL much lower than what needed (say 2 seconds vs 20
minutes) that would result in memory-related issues.

If you use the SP_HDR with raw sockets, and are using msg_control, be
advised that the SP_HDR payload that is delivered (and processed) now begins
with a size_t indicating the size in bytes of the actual SP headers that
follow. You will need to update your code accordingly if you are using this
API.

Also a number of problems including races around nn_shutdown and nn_close
are now resolved.

This does not fix all known issues, and the quality is still not where I
think it needs to be to declare this production ready. But substantial
improvements in that direction have been made in this release.

Everyone who is able and running older releases of nanomsg should switch to
this release asap.

Will do!

Bests,
Stefano


- Garrett


Other related posts:

  1. Home
  2. nanomsg
  3. Archive
  4. 11-2015