Replaced boost by c++11 and asio

[chohner]

This completely strips Boost from the network interface:

• Include asio.hpp and define ASIO_STANDALONE
• boost > std: (enabled_shared_from_this, shared_ptr, bind,
  asio::placeholders, system::, thread)
• boost::asio > asio (ip::tcp::socket, ip::tcp::acceptor,
  ip::tcp::endpoint, io_service, error_code, streambuf, async_write)
• boost::asio::deadline_timer becomes asio::steady_timer (to work with
  std::chrono)
• boost::posix_time > std::chrono

This should compile already after installing asio (brew install asio on OSX, apt-get install libasio-dev on unix).

Todo:

• Adjust build environment (check against asio instead of boost)
• Update readme on installing asio and enforcing c++11

[chris-hld]

Yep, compiles on Ubuntu 16.04.

[mgeier]

Cool, that works great for me on Debian testing!

All the ugly Boost stuff in configure.ac can now be replaced with those few lines:

ENABLE_FORCED([ip-interface], [network (TCP/IP) interface],
[
  CPPFLAGS_backup="$CPPFLAGS"
  CPPFLAGS="$CPPFLAGS -DASIO_STANDALONE"
  AC_CHECK_HEADER([asio.hpp], , [have_ip_interface=no])
  CPPFLAGS="$CPPFLAGS_backup"
])

I think it makes sense to define ASIO_STANDALONE in the code where you did it, therefore it should be un-defined again after checking for asio.hpp.

Or probably it would be even better to do something like this (and removing the CPPFLAGS_backup lines above):

#if !defined(ASIO_STANDALONE)
#define ASIO_STANDALONE
#endif

[mgeier]

Cool stuff, I've just added a little comment regarding the docs.

[mgeier]

Please don't remove this. Instead, add "This issue was resolved in ...".

[mgeier]

Please add the appropriate SSR version number.

[chohner]

What would be the correct version number? 0.4.3? Or would a larger version jump be more appropriate? Especially if this is combined with the upgrade of qt5.

[mgeier]

I think 0.5.0, but @JensAhrens will make the release and he will decide if this PR will be included or not.

[mgeier]

@chohner When you are done, can you please squash your commits?

[JensAhrens]

Yes, 0.5.0 is it!

[JensAhrens]

... and I'm happy to include this PR.

[umlaeute]

maybe the use of ASIO_STANDALONE should be made settable by the person who is compiling (i don't have any special use-case in mind though).

something like --disable-asio-standalone to be added to configure if you don't want it.

[chohner]

maybe the use of ASIO_STANDALONE should be made settable by the person who is compiling (i don't have any special use-case in mind though).

That switch would then add boost dependencies deep inside the asio library that you would need to tell the compiler/linker about and so on. I don't think there is a good reason for it anyways (as you also mention).

[umlaeute]

are these dependencies build-deps or runtime dependencies?

why does asio provide a non-standalone option (and even defaults to that) if it is not desirable to use it?

[chohner]

are these dependencies build-deps or runtime dependencies?

Build time.

why does asio provide a non-standalone option (and even defaults to that) if it is not desirable to use it?

I think only for historical reasons. The standalone option is still (comparatively) new and relies on c11. Since afaik there's plenty of c11 code throughout SSR, the only upside of building with boost (building with on std<c11) disappears. Additionally, some problems we were seeing on OS X also disappeared on this build so I don't think there's any reason to allow building with boost.

[chohner]

Small update from my side: I can confirm the interface to be working with the SSR android app.

[chohner]

Added a line on being fixed in version 0.5.0 and clarified the second line. The squashing can/should be done on your part when merging (squash&merge is available directly in the github GUI).
Feel free to use something like this as the commit description:

* replaced boost::asio by c++11 and asio
* cleaned up configure.ac
* rename src/boostnetwork to src/network
* update manual

[mgeier]

Thanks @chohner.

You squashing your commits (or somehow else pushing them into shape) places the burden of coming up with commit messages on you, that's all.

Anyway, @JensAhrens will merge it, so I don't really care.

[chris-hld]

Tested in a compound version (https://github.com/chris-hld/ssr/tree/beta) and it works great for WFS.

[mgeier]

Apparently, I've approved this too quickly. Creating the manual produces a few errors now.

[JensAhrens]

I fixed the manual.