GitLab

Now we trig the play of the voice letter after the completion of the
message 'click on letter'. Before there was a fixed 1500ms but the
message is of variable length depending on the locale.

Extended GCAudio with the signals Error and Done.

Added the error signal in GCAudio to let the caller catch it.

It makes QtCreator to recognize our class in the editor.
Tested on desktop and android the qrc: reference is not needed.

- now all images are svg
- auto detect if we are a carriage or a cloud
- better graphical effects

First version of KConfig handling. Audio is replaced by GCAudio (enables or not the audio), config is saved (only work for audio), a first config tab

First version of KConfig handling. Audio is replaced by GCAudio (enables or not the audio), config is saved (only work for audio), a first config tab

Answers and questions are loaded from a levels file to allow for
non-latin tasks. Names of levels files are expected in the form
levels-<locale>.json and must contain valid json describing the
levels (cf. the contained default levels-en.json).

Non-default levels files are supposed to live in some to-be-defined
datapath directory (for now faked). Upon loading through a File
object JSON syntax and basic payload checks are performed. If they
fail we fall back to the packaged default file: levels-en.json

ToDo / Open Issues:

- Sound with qrc source schemes not working on Linux
(see https://bugreports.qt-project.org/browse/QTBUG-36e724).
- No voices sounds (on linux) unless symlinked manually.
- "fx"/sound settings faked by getSettings(), need settings
integration.
- No scaling on Linux because ApplicationInfo.ratio is never updated.
- Levels are statically initialized. Allow to initialize from file
(for non-latin locales), probably needs file IO support from C++.
- Flicking the train on Android can be improved.