Point out get_n_children() for checking nullable status, and copy the
doc for Variant<VariantBase> to Variant< Variant<T> >.

https://bugzilla.gnome.org/show_bug.cgi?id=778219

Bug 786824

Due to an extra blank some files were not processed by gmmproc.

Modify the conditional inclusion of desktopappinfo.h in giomm.h and
wrap_init.cc. It shall not be included when running under MacOS. This should
have been fixed by commit fe99895f.
Bug 781947

Commit 0bc35fb4 fixed giomm_files_hg,
but in the wrong way. giomm_files_hg shall always include all .hg files,
regardless of which ones are used on the present platform.

plus some minor tweaks that don't quite seem worth changing upstream.

This adds some useful extra context for get_(prg|user|real)name().

create_certificate_handle_vfunc() shall delete the returned character array
after it has been copied to a Glib::ustring, or else it leaks memory.
Bug 783360

Dereferencing a null pointer does not necessarily cause a segfault:
it is undefined behaviour, which means anything, or nothing, may happen.
Avoid setting an expectation that a segfault will reliably occur here.

I cherry-picked a commit from master that should only have updated
documentation, but because the way of declaring slots has been changed
in libsigc++, it ended up changing them here too. That was unintentional

Daniel Boles authored May 29, 2017 and Daniel Boles committed May 29, 2017

This looks much less ugly.

Remove the default value of the 'type' parameter. As there is another
create() with one parameter, the default value makes a call to
create("pathname") ambiguous. Bug 782592

This fixes a build breakage, when building from a fresh checkout,
introduced by commit dd671f29 .

Bug 775741

 • Clarify why the old overload has been deprecated: it simply does not
   work for the desired result because the parameter_type was not passed

 • Elaborate on the purpose of the parameter_type in the new overload.

https://bugzilla.gnome.org/show_bug.cgi?id=774444

I replaced the old line that both created an Action without a parameter
and added it, only with a line that creates an Action with a
parameter... but did not add it. Of course, we need to do that, too.

https://bugzilla.gnome.org/show_bug.cgi?id=774444

U-eagle\daniel.boles authored Feb 02, 2017 and Murray Cumming committed May 04, 2017

If I’d realised there was a specific commit that protected this, I would
have reverted it, rather than doing it manually in my previous commit.

Completes reversion of commit b7431bca –
“Gio::SimpleAction: Make set_state() protected.”

https://bugzilla.gnome.org/show_bug.cgi?id=777953

Daniel Boles authored Jan 30, 2017 and Murray Cumming committed May 04, 2017

This is needed to apply the requested state in a custom handler for
signal_change_state(). It was protected on an assumption that when the C
doc for g_simple_action_set_state() said “This should only be called by
the implementor of the action”, it meant the implementor of a subclass.

But the C docs for signal change-state are clear that it means the
implementor of the signal handler. Without such a user being able to
call set_state(), they can’t apply the new state, unless they subclass
SimpleAction, which is a lot to require for such a simple case as this.

https://bugzilla.gnome.org/show_bug.cgi?id=777953

To avoid breaking ABI due to this commit that added a parameter:
https://git.gnome.org/browse/glibmm/commit/?h=glibmm-2-52&id=19ea8885ef17743c7c6e436e770b3b3835ca5016
Bug #774444

Daniel Boles authored Feb 13, 2017 and Murray Cumming committed May 04, 2017

This matches e.g. simpleaction.hg and is what I originally intended.

https://bugzilla.gnome.org/show_bug.cgi?id=774444

Daniel Boles authored Dec 24, 2016 and Murray Cumming committed May 04, 2017

This puts several things in more logical places and adds a couple of
line breaks that make it easier to see the key sections in the header.

https://bugzilla.gnome.org/show_bug.cgi?id=774444

Daniel Boles authored Dec 24, 2016 and Murray Cumming committed May 04, 2017

It was not registering the parameter type for the action on create, so
when activating the action later and passing a parameter, an assertion
failed about the expected and passed parameter types/counts not matching
and the parameter was not usable.

https://bugzilla.gnome.org/show_bug.cgi?id=774444

Daniel Boles authored Dec 01, 2016 and Murray Cumming committed May 04, 2017

https://bugzilla.gnome.org/show_bug.cgi?id=774903

John Ralls authored Apr 29, 2017 and Murray Cumming committed Apr 30, 2017

Disable building desktopapinfo.cpp with the same test used in glib.

g_object_newv() is deprecated in favor of g_object_new_with_properties(),
which is new in glib 2.54.

This is necessary for C++17 which completely removes dynamic
exception specifications.
See https://bugzilla.redhat.com/show_bug.cgi?id=1438766

This is necessary for C++17 which completely removes dynamic
exception specifications.
See https://bugzilla.redhat.com/show_bug.cgi?id=1438766

This is the first 2.51.* version of glibmm-2.4.
There were previously some 2.51 versions of glibmm-2.52, which
installed in parallel with glibmm-2.4,
but there will not be any stable version of glibmm-2.52.
There might instead be a stable version of glibmm-2.54 eventually.

Bug 779936

In case anyone is as easily confused as me…

https://bugzilla.gnome.org/show_bug.cgi?id=778575

Clarify what is set and returned, and use lower case “nothing” to match
how that must be written in GVariant text format, à la GSettings, etc.

https://bugzilla.gnome.org/show_bug.cgi?id=778219

It’s not exactly intuitive why the *_maybe() methods are kept in
ContainerBase, nor that Variant<VariantBase> inherits from the former
and is the way to get a maybe-typed Variant in glibmm. Let’s fix that!

https://bugzilla.gnome.org/show_bug.cgi?id=778219

Donʼt use braces for single-line blocks, and do if the other side of an
if/else already did. Also, get rid of a couple of extraneous newlines.

https://bugzilla.gnome.org/show_bug.cgi?id=778219