opengrok

• API used in Python tools now properly reports HTTP errors
• handle API error in Docker properly

• bump history chunk size
  • currently this is 128k changesets for Mercurial and 64k changesets for Git
    • the new historyChunkSize indexer tunable can be used to override this globally. Generally this should be needed only when working with very large repositories (think millions of changesets or high number of changesets that touch a lot of files) or when trying to drive the heap usage down radically (at the cost of increased indexing times) or when indexing many sizable repositories in parallel (the higher parallelism level the higher heap requirements). For the tunable, the semantics is: the higher the number, the higher memory consumption (depending on the size of the repository, i.e. number of changesets and the number of files touched by each changeset) and speed will be (assuming there is enough heap space available).
• Lucene 8.9.0
• per user settings
  • currently allows to disable the suggester

• better parallelization of the creation of history cache
  • the indexer option --historyRenamedThreads was renamed to --historyFileThreads
  • if using the -G indexer option to generate the tags, it is necessary to regenerate history cache from scratch (remove the historycache directory under the data root and re-run the indexer)

• webapp fixes

• smaller resource conservation fixes for the web app

• fix opengrok-mirror behavior when reading empty YAML file
• fix on the fly xref generation

• it is now possible to configure mirroring in Docker image
• opengrok-mirror has 2 new boolean configuration properties:
  • incoming_check - allows to override incoming check
  • ignore_errors

• Mercurial history is now generated in chunks
  • this is generally slower however has lower memory requirements
• JGit fixes

• fix Docker with SYNC_PERIOD_MINUTES: '0'

• history cache is now generated in chunks, lowering memory requirements significantly for repositories with big history
  • as a result, the history cache for repository top level directory is no longer supported
    • the OpenGrokDirHist.gz files in the historycache directories can be removed
• all indexer/web app Git code converted to JGit

• various Docker enhancements:
  • SYNC_TIME_MINUTES was renamed to SYNC_PERIOD_MINUTES
  • add CHECK_INDEX to force reindex from scratch on incompatible index change

• RESTful endpoint for Docker image to trigger reindex
• Docker image can now work with read-only configuration
• various small UI fixes

• fix searchall query parameter for sort options (prevents long URLs)
• another round of JGit conversion (bulk history)

• Lines/LOC are now accumulated for directories

NOTE: even though this release does not require reindex from scratch in the classical sense (i.e. wipe out data root), it basically updates every document in the index automatically. This means the initial reindex after upgrading to this version will take much longer. Subsequent reindex times will be normal.

• stale/missing xref are now generated on the fly
• restrict XML decoders

• CUDA files are now classified as C++ files
• TCP reachability of the web app is now tested when using the -U indexer option

• minor fixes (correctness)

• JGit fixes

• speed up xref redirections by storing the last revision of a file in the index
• more search.latency categories (monitoring)
• honor history setting in projects API call handling
• speed up Mercurial check
• first round of JGit changes
  • should speed up Git repository scan (indexer) and invalidation (indexer and web app)
• add the mergeCommitsEnabled tunable to disable handling merge commits
  • so far for Git only
  • provides workaround for high memory consumption
  • on by default
• -h guru indexer option now prints list of analyzers first
• INDEXER_OPT can be empty for Docker

• Docker: INDEXER_OPT is now honored correctly