* Fixed TaskFillIndexerCommandQueue exiting right away when all commands fitted into queue on first call
* Fixed TaskBuildIndex exiting right away when starting of indexer processes takes longer than the update cycle
* Refactored Indexing Tasks Behavior Tree
* Pass NameHierarchy to record new symbol returning an Id
* Cache symbol Ids in Parser implementation
* Reuse symbol Ids as reference when recording related data
* Pass FilePath to record new file returning an Id
* Cache file Ids in CanonicalFilePathCache
* Store file Id in ParseLocation instead of FilePath
fortune cookie message = You know a silent way to impose your will.
* Removed storage checks and single use methods
* cache file ids in ParserClientImpl to avoid lookup for every source location
* NameHierarchy::serialize with range to avoid copies when adding nodes
* removed duplicate code in IndexerCxx and IndexerJava, moved to Indexer
* use unique_ptr for CxxName handling
* create CxxNameResolver from other
* improved string handling in CxxName types
* only traverse decls within projects
* use macro in ast visitor to call into each component instead of virtual calls
* improved CxxName caching performance
* refactored CxxNameResolver classes to use less locals
* removed paths in include directives
* changed cmake to provide necessary header search paths
* changed name of version.h to productVersion.h due to name conflict
* moved SystemHeaderSearchPaths and FramewordSeearchPaths of indexer cxx indexer commands to CompilerFlags section
* added SourceGroupTestSuite to test indexer command generation of all source groups
* removed "compile_commands.json" from gitignore, because a resource file for the tests has this name
* fixed case where "to-index" state of a file has changed and thus the file needs to be re-indexed
* clear and re-index files that depend on changed non-indexed files
* clear and re-index files referenced by unchanged but cleared (due to referencing a changed file) files
* fixed: updated files with changed timestamp that do not have content in the db will now count as changed
* refined refresh info generator tests
* updated checklists for manual project setup tests
* implemented task for sending IndexerCommands to shared memory on demand
* SourceGroups now return IndexerCommandProviders instead of just a list of indexer commands. This way information can be stored in a compressed format
* merged IndexerCommandCxxEmpty and IndexerCommandCxxCdb
* moved sorting of indexer commands by file size from IndexerCommandList to TaskFillIndexerCommandQueue
* removed obsolete IndexerCommandList class
* wait for IndexerCommandQueue to be filled before starting the indexers
* restart finished indexer processes while indexerCommandQueue is still running
* restart indexer threads as long as there are indexer commands to process
* removed Blackboard::getMutex() and added an atomic update() method
* we don't need to get the complete compile commands for this
* as a fallback, querying a single compile command for a file is fast, because it is stored in a map
* use std::string instead of std::wstring in temp indices
* use number types for edges and source locations in temp indices
* use map::emplace where appropriate
* moved 100 occurence insert optimization fully to sqlite index
Eberhard Graether