The code should act the same as the old event handling. The code has
been lightly tested and can be disabled by defining
GUI2_OLD_EVENT_HANDLING during compilation.
The old code is still fully there and shall be moved withing #ifdefs
soon. Also the event handler has theoretic support for nested event
handlers but this code hasn't been used yet, which will also happen
later on.
...has been renamed to <preferences>/data/add-ons in order to better
reflect its current purpose: holding not only singleplayer campaigns,
but also any other kinds of single or multiplayer add-ons.
(Per mailing list thread,
https://mail.gna.org/public/wesnoth-dev/2009-03/msg00068.html).
2009-03-01T21:04:58Z!jeremy.rosen@enst-bretagne.fr reintroduced the old parts I just removed...
...and 2009-03-01T21:40:58Z!crazy-ivanovic@gmx.net removed those parts including some new parts again.
...is now the only titlescreen background again.
The old titlescreen files were moved into story/ and renamed as well;
wmllint has been updated to update existing content due to these
changes. Fixed mainline WML referencing those files, and added a release
note along with a quiet changelog entry.
...when loading images. Now all of them are C++ functors, and, except
for ~RC()/~TC(), they are accumulable and applied in a left-to-right
order from WML/C++.
This seems to have little, if any, performance impact in -O0+debug,
-O2+profiling and -O3 builds for me compared to the former
implementation. A test with gprof reports in average 0.0 seconds spent
in image::load_image_sub_file() for both implementations unless I'm
misinterpreting the data.
Tested in an AMD Athlon X2 Dual-core QL-62 on Debian lenny, non-stock
kernel 2.6.28.1 (SMP, model optimizations) running on native 64-bit mode
with the first scenario of HttT and a lot of hyperactive gryphons and
undead. 2 GB of RAM, 3.74 GB of swap. The compiler is GNU g++ 4.3.2.
...at badly mixed up paths when trying to locate UMC translations.
(If you like short stories, this just fixes bug #12872.)
(I've got no idea why get_binary_paths() is not delivering paths
with trailing slashes, but since this was the only victim, I
just did a quick hack in the textdomain initializer.)
The new dialogs didn't honour the escape action, which didn't show the
following messages. This behaviour is now implemented in the new widgets
as well.
The old style dialogs get dismissed on a mouse down, the next mouse up
event was most of the time handled by the new dialog. The new dialog
issued an easy_close event which closed the dialog. Moved the handler in
the mouse_up code (which already ignores mouse up events without the
button being down). The effect was percieved as the new dialog not
showing up.
