If the server keeps sending replies considered from the client in the
past (there is no way to tell if the server or client clock is wrong),
the client will try to interpolate a valid expiration time based
on the delta of the expiration intervals sent by the server.
If the intervals are always the same, exponential backoff kicks in
until we get a different expiration time from what we already have.
When looking for child tiles that cover a parent tile, we're computing the lower bound of the child tile in the set of tiles. Three things can happen:
- all remaining tiles are sorted before the child tile we're looking for. in that case, there can not be a child tile.
- the lower bound is the child tile; in this case, we want to continue checking the other children of the parent tile.
- the lower bound isn't the child tile we're looking for, but there are still tiles sorted after the tile we were looking for. these could be four children of the child tile we're looking for, so we're recursively descending.
what the current algorithm didn't take into account so far is that tiles with a higher w number (i.e. tiles that are outside the "main" tile pyramid of 0/0/0) were always sorted *after* the current w number. this meant that we thought there were still remaining children to be checked, so continued our recursive descent until eventually the number overflowed and we aborted our search. unfortunately, on Android this recursive descent and accompanying overflow caused a crash.
this patch fixes this by making sure that before we're continuing to descent, we're sure that the w numbers match: all tiles with a different (= higher) w number essentially mean the same as "all remaining tiles are sorted before the child tile we're looking for" for the purpose of finding covering children
This eliminates the reliance on ThreadContext to provide GLObjectStore, and statically enforces that GL cleanup functions happen only when GLObjectStore::performCleanup is called.
With the elimination of the Map thread, this becomes important because there may be multiple GLObjectStore's per-thread, and Map will need to ensure that the correct context is active when calling GLObjectStore::performCleanup.
Texture pool now allocates a block of 64 textures per time. Each block
contains a TextureObject which handles GL object lifetime. Blocks can be
added or removed according to the following rules:
1. If pool is empty, created a new block upon first call (lazy load);
2. Iterate blocks to find an available texture ID:
- If found, remove that from the list of available IDs for that block
and return its ID.
- Othewise, allocate a new block and grab an available ID from that.
3. When releasing a texture ID, check if the associated block is free.
If no IDs are currently in use, then abandon its TextureObject and
remove the block from the pool.
When inserting an cached resource, or removing a region, remove least-recently used resources and tiles, not used by offline regions, until the used database size, as calculated by multiplying the number of in-use pages by the page size, is less than the maximum cache size minus 5 times the page size.
In addition, OfflineDatabase may be configured to ignore cache puts of individual resources larger than a certain size.
This policy is similar but not identical to the former SQLiteCache policy:
* It accounts for offline, by exempting resources required by offline regions from eviction.
* It must delete from two tables (resources and tiles), rather than one. Currently the strategy is naive: evict 50 rows at a time from each table.
* It makes maximumCacheSize and maximumCacheEntrySize completely independent. The SQLiteCache implementation evicted when `usedSize > maximumCacheSize - 2 * maximumCacheEntrySize`. This evicts when `usedSize > maximumCacheSize - 5 * pageSize`.
* It uses a non-unlimited default value for maximumCacheSize: 50 MB. We should have always had a limit in place; "a cache without an eviction policy is a resource leak".
This reverts commit 6709bdcacd.
There was an issue with the script that removes '-lwebp' from WebP
linker flags, since we're statically linking. This is now fixed.
Moved EdgeInsets to geo.hpp so CameraOptions and Transform can refer to it. Added a padding option to CameraOptions that alters the frame of reference for the center option. Added optional padding parameters to LatLng getters and setters.
Working towards #2600.
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.
* Move asset:// URL handling to DefaultFileSource.
* AssetFileSource implements FileSource interface and follows familiar implementation patterns.
* Move default implementation to platform/default, zip implementation to platform/android.
* Don't bother with modified / expires / etag -- assets are not cached so it doesn't matter.
* Don't bother with interleaving individual IO calls on the implementation thread. That adds a lot of complexity for very little benefit.
* Standardize on std::exception_ptr as the error representation
(fixes#2854).
* Don't format textual strings at the error source; pass on the
constituent data via observer method parameters instead.
* Use the null object pattern to simplify observer notification code.
* Further refactoring for ResourceLoading tests.
