I didn't use 'fixes' because this doesn't strictly fix that specific
bug (I think?). Anyway, it does remove strdup which is mentioned in the
issue.
Also fix elem.free not allowing DEFAULTFOO_PT_STUFF and similar to
be freed.
event.register(event.mousedown, function(...) print(...) end)
event.unregister(event.mousedown, somefunc)
mouseclick event split into mousedown, mouseup, mousemove, mousewheel
keypress event split into keypress, keyrelease, textinput. key* events only contain keycode and scancode, don't attempt to represent a letter (was very broken at this before). Also have helpful shift/ctrl/alt flags passed in. textinput just represents inserted text, can probably even handle foreign characters.
register_step replaced with event.tick event
All legacy register_* and unregister_ functions are removed. There is a compatibility lua script, might embed it later. tpt.set_shortcuts / tpt.test also removed. event.getmodifiers added, just a misc function to get the currently held modifiers
Lots of code duplication to handle each event is removed, it's not handled in a more generic way. Although the Event class / child classes could use some work.
Still currently in process, there are some issues:
Windows version doesn't work, mac version might not work, opengl might not work
Icon doesn't work (on Linux at least)
Lua will need some changes, there are some sdl 1.2 hacks in there
When entering fullscreen, the window loses focus
When holding down mouse out of bounds, mouse move events stop being sent
When letting go of mouse out of bounds, mouseup event doesn't take into account double scale mode
Clicking on startup without moving mouse will draw at 0,0 for a frame
Renderer probably won't compile because USE_SDL doesn't entirely work
... and maybe others
Some nice things were done though:
no more blit2, sdl can do the scaling itself
3d effect removed, no reason to support this joke any longer
No need to support copy/paste ourselves, sdl does it now
text handling done much better now, separate events for key presses and text input
when a new window is shown, all events ignored until next tick (ignore textinput event if window shown from key press event like console)
Window scale can now be anything between 1 and 10 (suggest something other than 10 for maximum scale?). This required a number of subtle changes:
* made blit2 (PowderToySDL.cpp) handle scale modes correctly (it really only handled scale:2 correctly before)
* replaced `bool scale` with `int scale` everywhere in the options view/model/controller
* replaced the _large screen_ checkbox with a _window scale_ textbox in the options view
The new scale is only checked and applied when the options view is closed. There's no reason to not apply it live, I just chose not to.
This commit does *not* make TPT able to figure out an optimal scale mode at first run. It still suggests using scale:2 if it makes sense though.
I had doubts about using a second loop in blit2 but it doesn't seem to be an issue. If there's a more optimal way of going about what blit2 does, I haven't figured it out.
(Sublime seems to have eaten a few trailing spaces, hence there are a few seemingly identical pairs of lines in the diff.)
remove 'r' record shortcut
add tpt.record function. Still gives the user a confirm prompt
recordings now go into recordings/<timestamp>/, where timestamp is the time the recording was started. <timestamp> is returned by the tpt.record function. Each new recording starts the filenames over at 0 again.
you probably still need a lua script to use the recording feature, this should make it easier for those
`tpt.parts` does check whether the particle ID it gets is valid, but it doesn't check whether that particle ID is used. One could potentially modify the life property of dead particles to break the linked list of free particle IDs, thus potentially gaining the ability to read from or write to arbitrary addresses in memory.
The logging is saved inside the bson data in all online saves, local saves, stamps, and clipboard pieces. It is loaded back when reloading each of those.
See #474 for the format of the data. It is the same format for the bson data. Note that "links" is an array of objects. It can be recursive.
There is some effort to not duplicate information, we don't care if you loaded a stamp 10 times or if you are using the clipboard in your own save. Extra information is mostly not saved for your own stuff, only when you take material from other saves.
Press ctrl+a in debug builds to show what info it is currently saving in Client. Also enabled in snapshots for now.
There is one unrelated change in here, which fixes a crash pointed out by QuanTech. It was also save related and it was too close to the other changes to separate it into another commit. It fixes a crash when saving signs with invalid unicode. the BSON library doesn't like this, it was returning an error but we ignored it, which caused a crash. It now notices those errors. I also had to update several Serialize calls to check if it actually returned save data, or else it then would have started crashing there instead.
Also some debug prints were removed
