Summary:
Using a type visitor scheme similar to type and marshall_type, to ease
increasing coverage for other types if needed.
Also, changed the filter functions for those selections to check for the
full name of the regular_type_def, allowing a better specificity of the
name selections. This fixes things like "Eina.Error", which was
conflicting with Efl.Image.Load.Error when we compared only the last
name. It didn't appear before as Load.Error only appears in a
Efl.Image.Load event.
Depends on D5996
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5997
Summary:
While the test events were declared as by value, the support code passed
by reference, which was - wrongly - the only implementation supported in
event code.
Also added test with Eina_Bool and structs (passed by pointer to events,
while not having the ptr modifier explicitly.
Depends on D5995
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5996
Summary:
Making it easier to share code between self and inherited events.
During this move, the namespace and keyword headers were merged into the
name_helpers header.
Also added the first seed of a generic namespace reducer function,
to be used by other functions in later commits.
Depends on D5994
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: segfaultxavi, cedric
Differential Revision: https://phab.enlightenment.org/D5995
Summary:
Don't try to lazily create the parts when getting them, as it would
modify the owner inside its @const Efl.Part.part.
Depends on D5993
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5994
Summary:
Blacklist functions are all in the header/namespace blacklist.
Helper functions returning strings (names) are in the name_helpers
header. They act somewhat like "mini-generators".
Helpers.hh was left with other kind of helper functions (checks, etc)
that do not return strings.
Depends on D5992
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5993
Summary:
To deal with events with the same name as some methods (Del, Invalidate,
etc), the suffix Evt was added.
Thus, now we use
obj.ButtonClickedEvt += callback;
Instead of
obj.BUTTON_CLICKED += cal
The argument classes use the same scheme, being called <Evt name>_Args.
Depends on D5991
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5992
Summary: It was a reminiscing bit from the c++ generator.
Reviewers: felipealmeida
Reviewed By: felipealmeida
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5991
Summary: remove the elm legacy name of efl ui theme
Test Plan: run elementary_test and test efl ui widget cases
Reviewers: Jaehyun_Cho, woohyun, cedric, raster, jpeg
Reviewed By: Jaehyun_Cho
Differential Revision: https://phab.enlightenment.org/D5934
Summary:
the win has no theme set, so this would always return null
Depends on D5956
Reviewers: cedric
Reviewed By: cedric
Differential Revision: https://phab.enlightenment.org/D5957
there were some cases where frame geometry was being calculated/applied
strangely as a result of moving all of the layout calcs to pre-render.
enforcing a frame calc+resize resolves these issues
Differential Revision: https://phab.enlightenment.org/D5961
this should just be handled in the pre-render callback where the rest
of the calc for the window is done
also removes an unnecessary smart calc
Differential Revision: https://phab.enlightenment.org/D5960
this avoids a substantial number of unnecessary recalcs and halves the calltime
for elm_win_add
ref T6884
Differential Revision: https://phab.enlightenment.org/D5944
Summary:
further optimize the eet_shutdown case of closing files and avoid
endlessly looping over empty nodes
Depends on D5950
Reviewers: cedric
Reviewed By: cedric
Differential Revision: https://phab.enlightenment.org/D5951
no need to free mempool memory when the mempool is about to be freed,
and eet_clearcache duplicates existing shutdown functionality
Differential Revision: https://phab.enlightenment.org/D5950
neither of these functions should force a recalc under any circumstance
as they are simply returning pointers
@fix
Differential Revision: https://phab.enlightenment.org/D5947
a number of these functions have implicit recalcs, which is bad because
it's a pretty significant perf bottleneck, but it can't be improved
without breaking existing behavior expectations so this is probably the
best that can be done
ref T6884
Differential Revision: https://phab.enlightenment.org/D5946
We want to make sure to have a stable and reliable subset of builds that
define if the build passed or not.
We also want to have builds which are more experimental. They give us a
good insight, but we are not yet ready to have them supported officially
as need-to-pass build. Namely the macOS build.
Another side effect of this change is that we reduce the critical build
matrix to 5 builds. The exact number of parallel ones we are allowed. With
fast_finish Travis will not wait for the other ones to finish before
setting the build status. This will allow us to have all builds in
parallel and not waiting for build #6 to be finished.
The file names are derived from the test case names. These two seem to
have been overlooked when copying one test case to another.
I hoped this would fix my problems I have with the eio test suite
recently, but sadly it does not. Time to go on a git bisect tour.
Summary: Need to check whether Evas_Public_Data is null or not before dereferencing it.
Test Plan: Execute test suite
Reviewers: raster, Hermet, cedric, jpeg, stefan_schmidt, Jaehyun_Cho
Differential Revision: https://phab.enlightenment.org/D5987
Reviewed-by: Cedric Bail <cedric@osg.samsung.com>
a wildcard gfx object could bring focus to a evas object that has no
smart parent, and thus keyboard events are left at a place where they
cannot come away from.
clock was using this to give focus to the edit buttons. The edit buttons
are now wrapped into theire own adapter objects which makes them
focusable like before. However, the no keybindings or anything are taken
to react on the value.
the manager objects are build on the assertion that registered elements
are returning the manager they are registered on if
efl_ui_focus_object_manager_get is called.
If we don't ack_configure here, it'll just automatically happen in
ecore_wl2_commit() next time we render anyway.
If we do ack_configure here, we can totally screw up E's internal window
handling.
Removing it seems like a win.