Boost logo

Boost :

Subject: Re: [boost] [asio] Bug in boost asio on mac
From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2015-03-23 12:21:49


On 23 Mar 2015 at 15:03, Joël LAMOTTE wrote:

> > > I have had this bug on mac os Yosemite where I would call async_connect
> > and
> > > my code would never be called back. I didn't have any answer to my bug
> > > report and I would like to be sure that it it is not lost.
> > >
> > > https://svn.boost.org/trac/boost/ticket/11070
> >
> > Already fixed in ASIO standalone I believe. He got a bit out of sync
> > with hidden visibility markup for a while there.
>
> I am working with Philipe on this issue. Do you remember which branch/tag
> should have the fix?

The fix of it not working on OS X was in the changelog.

Also, in the day job we have latest standalone ASIO working as well
as it usually does with the kqueues reactor on FreeBSD and OS X. So I
know it's working.

> We didn't find it by just reading the code in master and are in the process
> of integrating github's asio
> see if it actually works, but we are having issues with converting asio to
> boost.asio (strangely the boostify
> script don't seem to work for us).
> We think it might not be a good idea to use master at the moment and maybe
> we should
> try a 1.10 tag instead, if it is supposed to include the fix?

The ASIO in Boost 1.57 and 1.58 is the v1.10 series. ASIO standalone
is the v1.11 series, and may enter Boost in 1.59 or 1.60.

> > BTW it's best to report bugs in ASIO on the standalone ASIO tracker
> > on github. You can see turnaround within the hour there.
> >
> >
> Thanks for the info. Does it mean that the boost trac tickets for asio
> are never almost processed?

As a personal observation, since the git conversion the Boost bug
tracker has become less used for many Boost libraries. The github
issue tracker is much better integrated into git, as is all the other
tooling like Jenkins, Coveralls etc.

As ASIO is also standalone, and indeed the Boost edition trails the
standalone edition by up to a year, I certainly can see why Boost
Trac would tend to be under monitored. By the time anyone reports a
bug in Boost.ASIO it almost certainly was reported and fixed in ASIO
standalone months ago.

Having just said that, ASIO was broken with the kqueues reactor for a
scary long time (more than a year) before anyone reported it. It
suggests BSD/Mac is not a popular ASIO platform. Or that many BSD/Mac
users disable the kqueue reactor, not unsurprising given the quirks
it has.

Niall

-- 
ned Productions Limited Consulting
http://www.nedproductions.biz/ 
http://ie.linkedin.com/in/nialldouglas/



Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk