- Merge pull request Project-OSRM#889 from Project-OSRM/experimental/…

…cuke_datastore

- change of travel_mode is now indicated, ie. ferries, shuttle trains, etc.
- update car and foot profiles, add ferry mode
- support for maxspeed tags in the form of 'countrycode:zone type' for car profile
- support way surface, tracktype, smoothness tags in car profile
- ability to get multiple points from /nearest by using num_results argument
- use double precision floating point math to fix some rounding issues
- fixed a number of bugs in the compressed graph format
- use smarter caching in the test suite, speeds up re-runs.
- remove compile-time time stamps for less rebuilds
- replace boost::variant w/ mapbox::util::variant
- tighter C++11 integration
- enable partitioning on link-time optimization, good riddance GCC 4.7
- better checking for luabind compilation, fixes some OSX woes
- use object libraries in cmake

Potentially breaking changes:
- removed remaining GCC 4.7 support
- the server API has changed to include travel mode
- bumped minimum CMake to 2.8.8

This is the shortened change log:

- Announcement of waypoints, closes Project-OSRM#584
- U-Turns at Via nodes are now allowed (not yet visible in frontend)
- Extraction of road network roughly 10% faster than previous release
- No inconvient use of space in /tmp anymore
- Added a distinct tool to check if hsgr file is valid: osrm-check-hsgr
- Added more unit tests based on boost/test.
- Use multiple cores during sorting steps
- Windows port much more stable, builds with significantly less errors

- Fixed a rare segfault in path query
- Fixed a bug in turn restriction combination including overlaps and one-ways
- Fixed bug in osrm-components tool
- Added timings to a number of locations
- Added BUILD_TOOLS cmake flag, deprecated WITH_TOOLS
- Fixed selection of alternative route name
- Many, many bigger and smaller code changes

This is the shortened change log:

- Windows support: Fetch the latest develop branch build from http://build.project-osrm.org
- MinGW compiler support
- Parallelization now uses Intel Thread Building blocks, bringing feature parity to all supported platforms
- Significant query speed improvements, about twice as fast
- The distance table feature does not need any zoom level anymore. Finds best nearest edge automatically
- First compressed and succinct data structure to improve memory footprint
- Improved and faster heuristics for polyline generalization
- Fixes a resource leak on shutdown
- Shared-Memory parallel sorting where applicable
- Lots of smaller and bigger changes to the code base
- Lots of code improvements thanks to C++11 and static analysis tools
- Better control the number of used threads during extraction and preprocessing
- Updated NodeJS bindings [2] supporting the new distance table API endpoint

hot-fixing an off-by-one error

This is the shortened change log:

- new API call: efficient distance table computation
- use 100 locations at max for dist table (for now)
- JSON output generation through a dedicated JSON container
- deactivate LTO on debug build
- fix performance regression in preprocessing
- use static assertions for better source code correctness
- add some static asserts to guard against memory usage regressions
- do less work when compressing geometries
- move most geometric calculations to float
- replace boost::unordered_map, move hash function for pairs into its own header
- refactor route name extraction into its own class, fix name extraction
- further includes untangling
- introduce range based for loops to traverse graphs
- further boost by STL replacements
- use lookup tables from some trigonometry functions
- fix compile regression in debug build
- further code improvements and bug fixes

Please not the following (minor) breaking changes:

- the JSON formatted output now omits empty, i.e. unused fields
- the response file names are now response.js(on) or route.gpx

gain a pretty big release with more than 150 commits:

- modernized code base to use C++11 language features
- decluttered the code base touching virtually every line of code
- cut down on static analyzer warnings
- much better linking performance on GCC 4.9
- fixed extraction on 32 bit systems (while officially unsupported)

bug fix release for a serious performance regression

This is again a pretty big release with more than 170 commits (adding…

… about 1500 LOCs). And it is the first release in a series to bring major improvements in memory usage. Depending on the size and location of the extract, you can expect 25--45% improvements.

This is the shortened change log:

- compress all road geometries (49% on the planet for cars)
- make all error messages JSON
- activated link time optimizations by default
- more test scenarios (25+)
- use forward decl's to cut compile times
- ongoing effort to make code more legible
- several lints fixed that were detected by facebooks flint
- several lints fixed that were detected by Coveritys static analyzer
- several lints fixed that were detected by OCLint
- remove remnants of C-Style
- add clang-format style file

The release is the result of hard work of all contributors and comes …

…at a hefty count of 278 commits. The shortened changelog is this:

- Fix performance regression in JSON generation
- Update and enhance test suite
- Activate link time optimizations
- Update speed profiles to be more realistic
- CMake now adds RPATH info to executables
- Refactored some code for better readability
- Removed pesky function ptr from CRC32 computation
- Enhanced feedback on command line options
- Escape characters in JSON instead of HTML-entetizing
- Decode URI-encoded requests
- Fix a possible path quality regression
- General support for non-X86(-64) targets, e.g. Raspberry Pi
- Remove remaining C-style function calls
- Fix minor GPX export issue
- Fix resource leak, coverity issue #1121924
- Generate pkgconfig file
- Compile and install libOSRM as static lib
- Lint removal in CMakeLists.txt
- Refactor perpendicular projection to nearest location
- Untangling includes to cut down compile time
- More precise line generalizations
- Make nearest neighbor queries more efficient
- And many, many more smaller and bigger changes

fix indentation