Even though this feature is currently unused we should use
a different library version for future development. While we
provide workarounds for the API the ABI has changed.

…c ALSA

calls.

This layer provides an API that is more consistent with the data used by RtMidi.
On the other hand it allows to optionally use a locking mechanism for all
sequencer calls so that ALSA may use one client for all threads for certain
tasks (or if it is requested).

The ALSA output devices should be still considered broken.

The ALSA output devices should be still considered broken.

# Conflicts:
#	RtMidi.cpp

MidiInAlsa::openPort
MidiInAlsa::getPortList
MidiOutAlsa::getPortList

Now, we can do some first tests. Test files follow…

This was broken as we use ALSA addresses to store port ids, now. These
are stored as unsigned values.

Signed-off-by: Tobias Schlemmer <keinstein@users.sourcforge.net>

…h virtual ports.

…losed.

WinMM does not support a good port selection scheme.
At least wine seems to ensure reliable port ids.

…uencer.

…tiy enum)

# Conflicts:
#	RtMidi.cpp

Modern versions of gcc and clang can compile files in a way that they
can detect bad access to memory as use after free or array overflow
when it happens, which improves the debugging of such errors.

… segmentation faults.

The client may be unable to get a write buffer for some reason.

… API detection.

Both changes are needed to provide virtual ports from the all API seclection.

This patch has been tested both with g++ 4.1 and 4.4.

…or-api

WinMM does not support a good port selection scheme.
At least wine seems to ensure reliable port ids.

This patch got lost between the different source trees.

…git into portdescriptor-api

It's up to the developers to deal with this case, anyway.

This patch changes the way error messages are handled:
• source code locations are stored separately in the exception. This allows
  a GUI to prevent confusion of the end user while maintaining the ability
  to debug RtMidi.
• all strings are marked with gettext_noopt so that it is easy to extract them.
• if RtMidi is compiled with gettext support, gettext is called in the constructor
  of the error message.

Drawbacks: This approach still doesn't support to use correct plural forms.
However, this limitation can be easily circumvented by adding a parameter
for the translated text.

Benefits: All calls to gettext are bundled in the library part of RtMidi.
This may prevent confusion that can occur when libraries like wxWidgets come
into play, which provide their own gettext implementation.

Missing parts: automatic initialisation of libintl.

…UI-friendly

…gettext support.

… exceptiions in a constructor.

…al tests.

The test has been performed with gcc.

…i18n+GUI-friendly

…hanges.

…it into i18n+GUI-friendly

…it into i18n+GUI-friendly

This also fixes some test cases, that had been broken by the last patches.

… construct std::shared_ptr

- Must use using x = y instead of typedef.
- std::shared_ptr's constructor accepting a pointer is marked explicit, unlike the Pointer class. So, I have put in explicit constructors everywhere.

- Had to change C++11 detection.
- Missing includes.
- Fixed a bug introduced in my previous commit wrapping something with Pointer<PortDescriptor>()

This enables real C++11 support (see pull request #2 on keinstein/rtmidi at github):

• Must use using x = y instead of typedef.
• std::shared_ptr's constructor accepting a pointer is marked explicit,
  unlike the Pointer class. So, I have put in explicit constructors
  everywhere.

Only tested on OS X – will test on Windows & Linux(ALSA) soon.

As a sidenote – I’d really like to see this all merged into the official
RtMidi, or at least as a "version 3" branch. You’ve done a good job of
cleaning up a lot of things.

So far objects could decide whether they want to be cleaned up by RtMidi.
But this was not used by cancelCallback and the rtmidiin destructor.
Both functions call userCallback->delete_me(), now.

For certain reasons we use rtmidi_gettext() in some
files instead of calling gettext() or using the macro _().
This change must be made permanent.

We must be aware that it may take some time until the changes here
can be merged upstream. In the meantime developers and users should
be informed to blame the right person for bugs.

This is a whitespace ignoring merge in order to bring i18n+GUI-friendly to the state of origin/master