Ganga

---

ganga/python/Ganga
- Performance improvements when handling large lists of GangaObjects
- Removed Ganga re-exec during launch
- `runMonitoring` now has default timeout of 5min (300s) to allow long jobs to correctly monitor and update subjobs on backends
- Fixed bug in IPython buffer flushing whilst typing multi-line functions
- Fixed problem with Registry file locking
- Improved shutdown code
- New interactive sessions will stop the monitoring loop if existing sessions are discovered via lockfiles on disk
*\* type `enableMonitoring()` to re-enable monitoring in these sessions if required (may be unsafe)
- Various code cleanups

---

ganga/python/GangaDirac
- Added `monitorJobs` and `getBulkStateTime` to Ganga<->(LHCb)Dirac interface
- Significantly reduced number of calls to DIRAC backend for job monitoring/completion
- Slightly better handling of missing/invalid DIRAC proxy whilst running
- Bugfixes around Offline file Splitter and optimisations around extremely large numbers of LFN (list->set)
- General performance improvements

---

ganga/python/GangaLHCb
- Fixing some bugs in the construction of large Datasets from BKQuery
*\* (most time taken in executing a new large request against DIRAC)
- Added intial (WORK IN PROGRESS) support for CMake projects via new `GaudiRun` Application
*\* Still doesn't fully support Gauss+GaussSplitter
*\* Initial functionality of old-style Gaudi and GaudiPython applications complete
*\* No support (yet) for MassStorageFile in the inputdata
*\* No support (yes) for unique inputfiles per subjob via a splitter
- Improved handling of LHCbDirac environment caching

---

ganga/python/Ganga
- Adding individual postprocessors to jobs has been fixed 402
- Condor jobs can split by dataset 390
- Minor monitoring fix when there are strange errors from the Condor backend
- Fixed some minor bugs in preparing some jobs
- GangaDataset made more user friendly

---

ganga/python/GangaDirac
- Dirac has gained the option to set the default output SE for files (optional)
- Some performance improvements with communicating with the DIRAC server (reduced Dirac calls/reduced code fluff in Ganga)
- Some older configuration settings have been deprecated, some new configuration options have been added.
Removed: DiracOutputDataSE, load_default_Dirac_backend, DiracDefaultSE
Added: RequireDefaultSE, statusmapping (for devs), queueable_dirac_statuses (for devs), userVO
- Can now use the bashrc directly from dirac-ui and elsewhere so don't need a Ganga specific env conig file
- DIRAC 'Completed' jobs are now considered to be 'running' due to the transfer of output data
- Fixed bugs where files vanished during jobs completing and jobs reverting status within DIRAC fixed

---

ganga/python/GangaLHCb
- All new DIRAC settings have been made configurable within LHCb config
- LHCb config now managed LHCb specific settings (RequireDefaultSE, DiracCommandFiles, userVO, allDiracSE)
- LHCb now inherits more closely from vanilla DIRAC in Ganga even though all code is executed against an LHCbDIRAC instance
- Added initial support for LHCb specific integration tests

---

ganga/python/GangaAtlas
- Ignore comments in files input for --inPfnListFile
- Added the output_loc_to_input option to AthenaSplitterJob to map input files to output dirs
- --inPfnListFile now takes a comma separated list and uses output_loc_to_input to
direct the different outputs to appropriate dirs
- Patches from HammerCloud to set homepackage correctly and use AGIS JSON file for queue/DDM mapping
- Some Atlas code cleanup

---

ganga/python/GangaLSST
- New plugin for the LSST experiment. DIRAC RTHandler finsihed so grid jobs work and have been tested.

---

ganga/python/Ganga
- This release has improved the resilience of Ganga to corruption in the on-disk data repository.
- The tasks system has been fixed and should now be fully functional.

---

ganga/python/Ganga
- This release has once more focused on stability and reducing thread-based deadlocks
- Some spurious warning messages on startup have been removed. PR 227

---

ganga/python/Ganga
- This release focused on making sure that jobs are flushed to disk in a timely fashion. They should all be flushed every 30 seconds as well as at the beginning and end of submit()
- job.subjobs.select(...) now works as expected. PR 253
- The messages printed by the monitoring loop's heartbeat checker have been temporarily disabled. PR 269

---

ganga/python/GangaLHCb
- A fix for the Gaudi bug where DataObjectDescriptorCollection() was not found. PR 280

---

ganga/python/Ganga

This release has focussed on the overall stability of Ganga in order to avoid data loss and lockups in the background threads.
Also situations where the status of jobs is accidentally reverted have been eliminated.
- Fix parent of jobs being set to a Descriptor incorrectly
- Fix some jobs not being picked up my the monitoring loop. Issue 85
- Cleanup of GangaObject and proxy internals to make things more streamlined and consistent. PR 163
- Fixes for stomp sometimes stalling. PR 171
- Fix for cases where two threads attempt to write the same repository file at the same time, corrupting `data`. PR 186
- GangaObject is now thread-safe for reads and writes to schema attributes. PR 191
- Added a heartbeat monitor for the monitoring threads to catch stalled tasks. PR 200
- Fixed most common cases of the monitoring locking up against the registry. PR 09
- Fix Job.copy() not copying all its attributes. PR 217