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".
There is no such thing as a cancelled response, only cancelled requests. A request that is cancelled does not have its callback called with a Response.
This allows the FileSource interface itself to support revalidation. We could (and probably should) now rewrite HTTPContextBase implementations as FileSource implementations.
I regenerated assets.zip so that all file paths have an `assets/` prefix, as the Android AssetFileSource implementation asserts, and removed `TEST_DATA` from the paths.
* 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.
Added aliases for std::chrono typedefs (eg. 'Seconds' for
std::chrono::seconds). These aliases are used together with templated
helper functions to replace time_t with std::chrono::seconds for most
cases, in particular for 'modified' and 'expires' values in Response.
We're now reparsing tiles when they expire. We're also swapping out buckets atomically to avoid flickering data; i.e. we're displaying the old data as long as we don't have a new parsed bucket for that layer yet. The parsed buckets now live in the *TileData objects rather than in the TileWorker; only partially parsed == pending buckets will remain in the TileWorker. Once they're parsed, they're moved to the *TileData object.
It's not implemented in GCC 4.9.2's stdlib (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57250). Instead, we're now always using a mutex to protect access; we previously created a mutex only on cancelation, but since we're always canceling now, it makes sense to allocate it right away.
We're now returning stale responses from cache. Those responses will have the `stale` flag set to true. Currently, all requesters in the core code discard stale responses, and cancel the request immediately after they got a non-stale response.
This still has the unwanted side effect off blocking the Map thread
waiting for the workers to return (because they hold a reference
to the Map thread main loop).
Maybe the only way to solve this would be having a I/O thread for
dispatching messages.
The Request is not bound anymore to the invoker thread because the workers
will request for glyphs, but the reply will arrive at the Map thread.
