Avoid unrotating due to programmatic rotation, even programmatic rotation that might expose empty portions of the map. Only snap to north due to a gesture.
Updating glyphs is still unsupported, and there's no good use case for doing so. When we're using a stale glyph PBF, and the fresh answer contains changed to that glyph, we will continue to use the old glyph.
Previously, we could first get a stale response from cache, then immediately a "connection error" response that overwrote the error state variable. In the tileLoadingCompleteCallback, we'd only see the second overwritten error state and abort. This could lead to the map sometimes not being rendered at all until the user interacts with the map.
This is a naïve implementation that essentially merges updated data into existing data. It will *not* remove icons from the stale sprite if they aren't present in the fresh sprite (we aren't tracking the source of a sprite, and the user could have changed it as well). Similarly, it will not update icons that have changed in dimension. This is a rare edge case and probably not worth implementing.
Otherwise, this will remove them from the map and overwrite the old data. This could happen e.g. when we have connection trouble and a tile needs to be refreshed, or we're displaying stale information and the connection is down.
We're now supporting using stale TileJSON and GeoJSON data. When we receive a new answer with an updated TileJSON file, we're replacing the Source's metadata with the new one and trigger updates to make sure we're loading the correct tiles. Similarly, GeoJSON data will be reparsed.
This adds support for using cached styles that are stale. They're treated like changing styles; when the refreshed style changed compared to the one we've already had, we're swapping out the entire style, which might cause a slight flicker.
If you use many different caches, expired weak_ptrs will pile up in the unordered_map, but that is an edge case, and you probably shouldn't do that anyway.
Previously, the placement of symbols depended on the order in which the
glyphs they need were loaded. The placement order should be based on the
style.
To fix this, symbol layers are placed in `placementPending` after they
are parsed. After all layers are parsed, symbol layers (stored in
`placementPending`) are placed and returned.
This is a tricky one:
We are storing a lambda at the member variable `transitionFrameFn` and
capturing `this` at this lambda function.
At some point when running the callback, we set `transitionFrameFn` to
`nullptr`, effectively destroying the callback that is being executed.
After this point, `this` is no longer valid, as the memory gets
cleared and it might point to the correct location (or not). That
is why it works on some compilers but not on g++ 5.3.1. Touching
any `this->[smth]` is undefined after that.
The fix is just move `transitionFrameFn = nullptr` to the last
thing we do on the callback. This behaves similarly to calling
`delete this` on a class method.
Cleaned up gesture recognizers, flipping the origins of anchor points only and always before calling mbgl methods. This exposed the fact that the cursor locking feature during drag-zoom/rotate/tilt was broken, so I fixed it to match MapKit behavior exactly and also account for modifier changes in mid-gesture.
All Transform methods that take a PrecisionPoint now assume a “flipped” origin at the upper-left corner of the view. Previously, some methods assumed an origin at the lower-left corner.
