Re: [liblo] New C++11 bindings available for testing

Now works with clang++ !

Tested with clang++ 3.2.1 on Ubuntu 13.04, and clang++ 4.2 on OS X 10.8.4.
On Mac, I get an error message on running cpp_test, looking into it.

Steve


On Tue, Aug 6, 2013 at 2:18 PM, Stephen Sinclair <rad...@gm...> wrote:
> I forgot to mention that I have only been testing these bindings with
> my Ubuntu 13.04 install.  I have just found problems trying to get it
> working on my Mac.
>
> Here is how you can get past "configure" on OS X 10.8.4:
>
> ./configure CC=clang CXX="clang++ -stdlib=libc++"
> ...
> checking whether C++11 lambdas are supported... yes
> ...
>
> Unfortunately it still results in a bunch of compilation errors with
> clang++, not sure if that's due to my own semantic errors or due to
> incomplete C++11 support.  I'll have to look into it.  In the meantime
> I recommend playing around with minimum version g++ 4.7 if you're
> interested in trying it.
>
> Steve
>
>
> On Mon, Aug 5, 2013 at 6:29 PM, Stephen Sinclair <rad...@gm...> wrote:
>> Hi everyone,
>>
>> I'd like to announce a "ready" version of the C++11 bindings for liblo
>> that I've been working on for some time now.  It is available in my
>> "cpp" branch on github.
>>
>> You can check out the "test" program here, which gives a bunch of
>> examples of how it can be used:
>>
>> https://github.com/radarsat1/liblo/blob/cpp/src/cpp_test.cpp
>>
>> In particular, it supports the following features:
>>
>> * Class-encapsulated access to liblo data structures (e.g.
>> lo::Message, lo::Server, etc...)
>>
>> * Assign any kind of "callable" as a method handler
>>   - normal functions
>>   - function objects ("functors")
>>   - lambdas
>>
>> * Method handlers may declare only the arguments actually needed
>>
>> * Method handlers need not have a return value, defaults to adding
>> "return 0" if your method handler returns void (return-type
>> polymorphism!)
>>  -- (this decision was based on the idea that "return 0" is most
>> commonly used.. not based on a scientific survey!  feel free to
>> disagree.)
>>
>> * Construct bundles and messages using static initializer lists
>>
>> * Polymorphism when adding data to messages
>>
>> This is all implemented in lo/lo_cpp.h, seen here:
>>
>> https://github.com/radarsat1/liblo/blob/cpp/lo/lo_cpp.h
>>
>> It's pretty complicated, requiring a bunch of macro-fu, so if you have
>> any questions don't hesitate.  (I would have preferred using only
>> templates but some of these tricks seemed impossible without help from
>> macros.)
>>
>> I think you'll enjoy using liblo through C++11.  Here's an example of
>> a method handler declaration:
>>
>>     s.add_method("test12", "i", [j](const lo::Message m)
>>                  {printf("test12 source: %s\n", m.source().url().c_str());});
>>
>> Here's an example of constructing a bundle and sending it in one line of code:
>>
>>     a.send(
>>         lo::Bundle({
>>                 {"test11", lo::Message("is",20,"first in bundle")},
>>                 {"test11", lo::Message("is",30,"second in bundle")}
>>             })
>>         );
>>
>> Enjoy, and please test, let me know what you think, etc.
>>
>> I've tried hard to make sure memory is handled right, but still
>> occasionally find problem cases.  It's a bit tricky and depends
>> heavily on the new refcount feature for messages and bundles. There
>> could definitely still be bugs but I have been exercising it quite a
>> lot.  Especially tell me if you encounter a "double-free" segfault.
>>
>> (For instance, one thing to know is that "lo::Message(lo::Message(m))"
>> constructs two wrappers around "lo_message m", but does not copy the
>> memory, they both refer to the same lo_message.  This means you must
>> use lo::Message::clone() to keep a message passed to a handler.)
>>
>> Steve