Skip to content

v5.13.0-rc.3

@TheMarex TheMarex tagged this 24 Oct 10:43
- Changes from 5.13:
  - Fixed regression in CH code that cause very long query times.

- Changes from 5.12:
  - Profile:
    - Append cardinal directions from route relations to ref fields to improve instructions
    - Support of `distance` weight in foot and bicycle profiles
    - Support of relations processing
    - Added `way:get_location_tag(key)` method to get location-dependent tags https://github.com/Project-OSRM/osrm-backend/wiki/Using-location-dependent-data-in-profiles
    - Left-side driving mode is specified by a local Boolean flag `is_left_hand_driving` in `ExtractionWay` and `ExtractionTurn`
    - Support literal values for maxspeeds in NO, PL and ZA
  - Infrastructure:
    - Lua 5.1 support is removed due to lack of support in sol2 https://github.com/ThePhD/sol2/issues/302
    - Fixed pkg-config version of OSRM
    - Removed `.osrm.core` file since CoreCH is deprecated now.
  - Tools:
    - Because of boost/program_options#32 with boost 1.65+ we needed to change the behavior of the following flags to not accept `={true|false}` anymore:
      - `--use-locations-cache=false` becomes `--disable-location-cache`
      - `--parse-conditional-restrictions=true` becomes `--parse-conditional-restrictions`
      - The deprecated options `--use-level-cache` and `--generate-edge-lookup`
  - Bugfixes:
    - Fixed #4348: Some cases of sliproads pre-processing were broken
    - Fixed #4331: Correctly compute left/right modifiers of forks in case the fork is curved.
    - Fixed #4472: Correctly count the number of lanes using the delimter in `turn:lanes` tag.
    - Fixed #4214: Multiple runs of `osrm-partition` lead to crash.
    - Fixed #4348: Fix assorted problems around slip roads.
    - Fixed #4420: A bug that would result in unnecessary instructions, due to problems in suffix/prefix detection
  - Algorithm
    - Deprecate CoreCH functionality. Usage of CoreCH specific options will fall back to using CH with core_factor of 1.0
    - MLD uses a unidirectional Dijkstra for 1-to-N and N-to-1 matrices which yields speedup.
Assets 2
Loading