OS X MacTypes.h/libgnomecanvasmm Rect problem workaround when COREAUDIO=0
git-svn-id: svn://localhost/ardour2/branches/undo@788 d708f5d6-7413-0410-9779-e7cbd77b26cf
creating a memory leak(?) that will go away with the transition of XMLNode* to
shared_ptr<>. A few bits toward restoring history from XML.
git-svn-id: svn://localhost/ardour2/branches/undo@779 d708f5d6-7413-0410-9779-e7cbd77b26cf
Save state basics, including adding PBD::ID to the delinquents. Compiles but needs to be tested (because I can't get the whole thing to compile on OSX due to the Rect problem).
git-svn-id: svn://localhost/ardour2/branches/undo@769 d708f5d6-7413-0410-9779-e7cbd77b26cf
Nuke Serializable in favor of Stateful. Got rid of some warnings with stub
code.
git-svn-id: svn://localhost/ardour2/branches/undo@767 d708f5d6-7413-0410-9779-e7cbd77b26cf
It compiles and runs and seems to still work. Still needed is the actual
serialization. Time to merge back to trunk so things can be tested by all.
git-svn-id: svn://localhost/ardour2/branches/undo@754 d708f5d6-7413-0410-9779-e7cbd77b26cf
global {solo,mute,record enable,metering} state commands. Same philosophy as
the MementoCommand but using only the appropriate state and not the entire
state of the session.
git-svn-id: svn://localhost/ardour2/branches/undo@718 d708f5d6-7413-0410-9779-e7cbd77b26cf
A couple more Command conversions. Now all that's left are the half dozen
or so global mementos.
git-svn-id: svn://localhost/ardour2/branches/undo@698 d708f5d6-7413-0410-9779-e7cbd77b26cf
Memento(Redo|Undo)Command has a noop for the undo or redo respectively, and
we don't need both before and after state. This is primarily useful for
drag start/finish callbacks, and really only makes sense where wrapped by
(begin|commit)_reversible_command (a composite command).
Also a few more "normal" MementoCommands.
git-svn-id: svn://localhost/ardour2/branches/undo@695 d708f5d6-7413-0410-9779-e7cbd77b26cf
All the obvious MementoCommand grunt work. Now there's some
add_undo/add_redo_no_execute sprinkled around where one is separated from the
other (e.g. in different callbacks) or perhaps even where there's only an undo
and no redo. Also some sigc-based undo/redo pairs that probably need their own
Command class.
git-svn-id: svn://localhost/ardour2/branches/undo@692 d708f5d6-7413-0410-9779-e7cbd77b26cf
This is the first swath of changes, replacing add_undo with MementoCommand
pattern, through most of the editor_mouse.cc file. However there were a few
places that weren't symmetrical that I need to think about. The question is
whether to tweak things so that they are symmetrical (add_undo paired with
add_redo*), or to allow Commands to not be undoable or not be redoable. Your
thoughts are welcome.
git-svn-id: svn://localhost/ardour2/branches/undo@685 d708f5d6-7413-0410-9779-e7cbd77b26cf
I just had an epiphany. I tried so many ways to make saving function name and
args work, it never occured to me that you could just as easily save undo
information as a pair of mementos, even in the Command-based structure we
agreed on.
Since many (read: almost all) existing undo commands take this form:
begin_reversible_command (_("change fade in length"));
session->add_undo (arv->region.get_memento());
arv->region.set_fade_in_length (fade_length);
session->add_redo_no_execute (arv->region.get_memento());
commit_reversible_command ();
We are already doing the save a memento before and after work. All we need to
do is instantiate an appropriate instance of MementoCommand. So the above
becomes:
begin_reversible_command (_("change fade in length"));
MementoCommand<arv_region_t, arv_region_memento_t> before, after;
before = arv->region.get_memento();
arv->region.set_fade_in_length (fade_length);
after = arv->region.get_memento();
session->add_command(arv->region, before, after);
commit_reversible_command ();
(With apologies for being too lazy to go look up what arv_region_t and
arv_region_memento_t are)
Note that the true command approach is still possible, and encouraged (both by
dictate and design).
git-svn-id: svn://localhost/ardour2/branches/undo@680 d708f5d6-7413-0410-9779-e7cbd77b26cf
Coding for undo/redo starts in earnest. Paul and I decided to go with a
standard gang of four Command pattern, with serialization. This overcomes the
terrible difficulties we were having with static type checking and the sigc++
approach. I'm adding the requirement that each command support undo,
simplifying undo/redo. NOTE that an important fallout here is that
Command::operator()() is the opposite of the old UndoAction::operator()(), i.e.
Command::operator()() is execute/redo, and Command::undo() is undo.
This commit is a reworking of the infrastructure, and won't compile until
creating Command subclasses for the various commands being performed. That is
primarily where you find get_memento and/or calls to add_(undo|redo.*).
git-svn-id: svn://localhost/ardour2/branches/undo@655 d708f5d6-7413-0410-9779-e7cbd77b26cf
renamed UndoCommand to UndoTransaction, and created new UndoCommand class and
its templated subclass SlotCommand. Those two are still in considerable flux.
git-svn-id: svn://localhost/ardour2/branches/undo@606 d708f5d6-7413-0410-9779-e7cbd77b26cf
