* Fetching missing state from the roomserver.
Whenever the syncserver receives an event from the room server that adds
state that isn't in the syncserver's local database it should fetch
those state events from the roomserver.
* Fix append
* Put comment back
* Comments
* s/addsStateEvents/lookupStateEvents/
* Fix spelling
* Include the stream position that a state event was added at in the current state tables
* Fix comment
* Review comments
* Add API for querying events by ID.
* Fix tense
* Start implementing federation ingress
* More stuff
* Hook up federation event receiving
* Handle the case where we are missing state
* Fix docstring and comments
* Fix infinite loop when printing unknownRoomError
* vendor: Add bimg image processing library
bimg is MIT licensed. It depends on the C library libvips which is LGPL
v2.1+ licensed. libvips must be installed separately.
* mediaapi: Add YAML config file support
* mediaapi: Add thumbnail support
* mediaapi: Add missing thumbnail files
* travis: Add ppa and install libvips-dev
* travis: Another ppa and install libvips-dev attempt
* travis: Add sudo: required for sudo apt* usage
* mediaapi/thumbnailer: Make comparison code more readable
* mediaapi: Simplify logging of thumbnail properties
* mediaapi/thumbnailer: Rename metrics to fitness
Metrics is used in the context of monitoring with Prometheus so renaming
to avoid confusion.
* mediaapi/thumbnailer: Use math.Inf() for max aspect and size
* mediaapi/thumbnailer: Limit number of parallel generators
Fall back to selecting from already-/pre-generated thumbnails or serving
the original.
* mediaapi/thumbnailer: Split bimg code into separate file
* vendor: Add github.com/nfnt/resize pure go image scaler
* mediaapi/thumbnailer: Add nfnt/resize thumbnailer
* travis: Don't install libvips-dev via ppa
* mediaapi: Add notes to README about resizers
* mediaapi: Elaborate on scaling libs in README
* Add API for querying events by ID.
* Fix tense
* Start implementing federation ingress
* More stuff
* Hook up federation event receiving
* Fix comments
* Comment on the order of the arrays
Avoid locking around db requests by only locking around active requests
and always creating an active request if there is none. A nice side
effect of this is that if many parallel requests for remote media come
in (a common case) then only one database query is made for the duration
of the query.
