* origin/engine-vs-velocystream: (91 commits)
removed useless include
removed JsonParser
removed JsonHelper
Fixed another failed nullptr handling in LogicalCollection cluster case.
Fixed undefined behaviour in rest error creation
Implemented updated of LogicalCollectionInfo. Right now this is only used in Cluster Mode
agency startup changed to bootstrapping
Mention howto control core files on linux.
shutdown fixed
Removed create / drop system collection from Cluster test. This is forbidden now
The logical colleciton now knows it's journalSize
Fixed a bug in collectionNameResolver. It now returns cid 0 for not found collections in cluster again.
pushing corrected unittest script and cluster startup scripts
adjusting startup scripts
missing source file added
we have a bootstrapping agency
votes are processed correctly again
Add module.context.createDocumentationRouter
Don't implicitely build the windows packages.
remove debug print
...
Conflicts:
arangod/Cluster/ClusterComm.cpp
arangod/GeneralServer/VppCommTask.h
arangod/RestHandler/RestBaseHandler.cpp
lib/Rest/VppRequest.h
* origin/devel: (105 commits)
Fixed failure tests in SkiplistIndex
Fixed a bug with reset of new skiplist index iterator
whitespace removal
micro optimization for key generator key tracking
GraphQL! Hype! 🎉
explicit
remove unused variable
explicit
Fix VS complaints
complaints
complaints
updated CHANGELOG
fixed issue #1937
Modified the SkiplistIndex lookup s.t. it does not build one large VPackBuilder containing all combinations of search values, but instead has an option to rewrite it in-place and just has to maintain the intervals.
ignore revision ids
fix destructuring of Externals in MERGE
make it code.
ignore primary and edge indexes during cluster restore
grunt build [ci skip]
semist.
...
* origin/devel:
Start adding a hybrid logical clock.
Handle replicationFactor and numberOfShards in arangorestore.
upgrade script for homebrew
Fix namespace stuff for VC++
Use standard GNU install dirs
This adds an option --default-replication-factor to arangorestore and
makes it more convenient to specify the number of shards for restored
collections.
Explanation: Different replicas of the same shard will have different
local collection IDs. Therefore we have to use the shard name
(collection name) to indentify collections in synchronous replication.
I have added a thorough description of events to the comments
in ClusterComm.h. This should enable everybody to do proper
error handling when using ClusterComm::asyncRequest and
ClusterComm::syncRequest.