As gnome@ is lacking active committers at the moment, transfer some of its
ports [1] up the stack to the desktop@ group, in hope that this way we get some
updates in as the set of people that "should feel responsible" grows.
As soon as gnome@ grows some committers again, this can (and should) of course
be reverted again.
[1] The list of ports chosen in this move consits of all the ports that are
required to build x11/kde5.
Approved by: portmgr (bapt)
Differential Revision: https://reviews.freebsd.org/D26362
Release 20.09.0:
core:
* Compability fix for Forms
* Fix fetching of Objects failing in some cases
* Fix clearing date in Annot setModified/setDate
* TextSelectionPainter: support glyphless fonts
* Splash: Don't try read past end of image
* avoid abort() on large memory allocation
* Fix memory leak on broken files
* Fix potential invalid memory read
* Small code improvements
qt5:
* Document TextAnnotation::inplaceAlign
* Make Annotation::setModification/CreationDate work on existing annots
* Be a bit more stubborn converting dates that come from xml
* Clean as many null characters from the end as possible when converting strings
glib:
* Add accessor functions for PopplerAttachment
* Deprecate PopplerAttachment GTime fields
* Deprecate PopplerDocument date properties
utils:
* pdftoppm: report error and exit if output file cannot be written
* Document that PDF-file can be '-' to read it from stdin
build system:
* cmake: Modern way to link against libpng, zlib and libtiff
* cmake: Remove stray support for lcms1 in pdftocairo
PR: 249060
Exp-run by: antoine
This update drm-current-kmod to be in sync with the latest Linux LTS
release.
aarch64 is also now supported for amdgpu and radeon.
vboxvideo is now added.
Reviewed by: x11 (zeising@)
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D26367
Clickhouse for FreeBSD updating no memory usage stats. In the case build without NDEBUG,
allocator use 4KB chunk, those dont account when freeing and MemoryTracking metric grow
up to ~6TB, after this clickhouse stop query processing with error "Memory limit (total)
exceeded: would use 6.24 TiB" due skipping calc allocation < 4MB
Patch still not upstreamed, obtained from:
https://github.com/ClickHouse/ClickHouse/issues/13859 +
https://github.com/ClickHouse/ClickHouse/pull/13869
reported and tested by: Vyacheslav (via Clickhouse telegram)
PR: 249205
Submitted by: olevole@olevole.ru (maintainer)
Obtained from: clickhouse repository
Changelog:
* Notable issues fixed:
- Fixes in Docker image configuration of Monitoring and Central
Management:
- As more and more users adopt our docker images, we’ve been getting
reports on gaps where these images don’t provide all the configuration
knobs the other artifacts do. This release exposes more environment
variables for configuring proxy support and certificate verification
mode for monitoring and central management , and for configuring
verification_mode
- Pipeline execution fixes to flushing and shutdown
- Since 7.2.0, a change caused terminating pipelines to not request
input plugins to cleanup before shutdown, which could cause leaks in
resources that weren’t freed during pipeline reloads.
- Logstash supports ordered execution for pipelines with a single
worker. A bug was found in this mode where the flushing mechanism
wasn’t working, preventing plugins like the aggregate filter from
working correctly. This has been fixed, and you can read the details
here:
* Consistent Fingerprinting:
- Our fingerprint filter is a popular solution to perform deduplication
of data in downstream systems like Elasticsearch, by computing a hash
value based on data from each event. Users reported that this filter
could produced different values for events containing the same data
since it didn’t ensure the order in which Hash Maps/Objects/Ruby Hashes
processed their key/value pairs. This has now been fixed.
* Updated JRuby to 9.2.13.0:
- The new JRuby release brings greater stability to its code
optimizations in multithreaded workloads and a fix to exception handling
in Windows environments, both issues that could affect our users. See
the JRuby release notes for more information.
* Plugins:
- Avro Codec - 3.2.4
- [DOC] Add clarifications on partial deserialization
- Fingerprint Filter - 3.2.2
- Fixed lack of consistent fingerprints on Hash/Map objects
- Kv Filter - 4.4.1
- Fixed issue where a field_split_pattern containing a literal
backslash failed to match correctly
- Elasticsearch Input - 4.7.1
- [DOC] Updated sliced scroll link to resolve to correct location after
doc structure change
- [DOC] Added usage example of docinfo metadata
- Http_poller Input - 5.0.2
- [DOC]Expanded url option to include Manticore keys
- Snmp Input - 1.2.5
- Updated snmp4j library to v2.8.4
- Fixed: support SNMPv3 multiple identical security name with different
credentials
- Fixed: multithreading problem when using multiple snmp inputs with
multiple v3 credentials
- Syslog Input - 3.4.4
- Refactor: avoid global side-effect + cleanup
- avoid setting BasicSocket.do_not_reverse_lookup as it has side
effects for others
- Jdbc Integration - 5.0.6
- DOC:Replaced plugin_header file with plugin_header-integration file.
- Rabbitmq Integration - 7.1.1
- DOC:Replaced plugin_header file with plugin_header-integration file.
- Elasticsearch Output - 10.6.2
- [DOC] Added clarifying info on http compression settings and behaviors
- [DOC] Fixed entry for ilm_policy default value
PR: 249108
Submitted by: Juraj Lutter <juraj@lutter.sk>
Changelog:
* Notable issues fixed:
- Fixes in Docker image configuration of Monitoring and Central
Management:
- As more and more users adopt our docker images, we’ve been getting
reports on gaps where these images don’t provide all the configuration
knobs the other artifacts do. This release exposes more environment
variables for configuring proxy support and certificate verification
mode for monitoring and central management , and for configuring
verification_mode
- Pipeline execution fixes to flushing and shutdown
- Since 7.2.0, a change caused terminating pipelines to not request
input plugins to cleanup before shutdown, which could cause leaks in
resources that weren’t freed during pipeline reloads.
- Logstash supports ordered execution for pipelines with a single
worker. A bug was found in this mode where the flushing mechanism
wasn’t working, preventing plugins like the aggregate filter from
working correctly. This has been fixed.
* Consistent Fingerprinting:
- Our fingerprint filter is a popular solution to perform deduplication
of data in downstream systems like Elasticsearch, by computing a hash
value based on data from each event. Users reported that this filter
could produced different values for events containing the same data
since it didn’t ensure the order in which Hash Maps/Objects/Ruby Hashes
processed their key/value pairs. This has now been fixed.
* Updated JRuby to 9.2.13.0:
- The new JRuby release brings greater stability to its code
optimizations in multithreaded workloads and a fix to exception
handling in Windows environments, both issues that could affect our
users. See the JRuby release notes for more information.
* Plugins:
- Avro Codec - 3.2.4
- [DOC] Add clarifications on partial deserialization
- Fingerprint Filter - 3.2.2
- Fixed lack of consistent fingerprints on Hash/Map objects
- Kv Filter - 4.4.1
- Fixed issue where a field_split_pattern containing a literal
backslash failed to match correctly
- Elasticsearch Input - 4.7.1
- [DOC] Updated sliced scroll link to resolve to correct location after
doc structure change
- [DOC] Added usage example of docinfo metadata
- Http_poller Input - 5.0.2
- [DOC]Expanded url option to include Manticore keys
- Snmp Input - 1.2.5
- Updated snmp4j library to v2.8.4
- Fixed: support SNMPv3 multiple identical security name with different
credentials
- Fixed: multithreading problem when using multiple snmp inputs with
multiple v3 credentials
- Syslog Input - 3.4.4
- Refactor: avoid global side-effect + cleanup
- avoid setting BasicSocket.do_not_reverse_lookup as it has side
effects for others
- Jdbc Integration - 5.0.6
- DOC:Replaced plugin_header file with plugin_header-integration file.
- Rabbitmq Integration - 7.1.1
- DOC:Replaced plugin_header file with plugin_header-integration file.
- Elasticsearch Output - 10.6.2
- [DOC] Added clarifying info on http compression settings and behaviors
- [DOC] Fixed entry for ilm_policy default value
PR: 249108
Submitted by: Juraj Lutter <juraj@lutter.sk>
Changelog:
* Breaking changes:
- Indices changes:
- Elasticsearch includes built-in index templates for logs-*-* and
metrics-*-*.
- Script cache changes:
- The script.cache.max_size setting is deprecated.
- The script.cache.expire setting is deprecated.
- The script.max_compilations_rate setting is deprecated.
- Mapping actions have been deprecated for the create_doc, create,
index and write privileges.
- Settings changes:
- Automatically importing dangling indices is disabled by default.
* New features:
- Search
- QL: Wildcard field type support
* Enhancements:
- CRUD:
- Log more information when mappings fail on index creation
- EQL:
- Make endsWith function use a wildcard ES query wherever possible
- Make stringContains function use a wildcard ES query wherever
possible
- Features/Stats
- Change severity of negative stats messages from WARN to DEBUG
- Search
- Fix handling of alias filter in SearchService
- QL: Add filtering Query DSL support to IndexResolver
- Snapshot/Restore
- Do not access snapshot repo on dedicated voting-only master node
* Bug fixesedit
- Authentication
- Call ActionListener.onResponse exactly once
- Authorization
- Relax the index access control check for scroll searches
- CCR
Relax ShardFollowTasksExecutor validation
Set timeout of auto put-follow request to unbounded
Set timeout of master node requests on follower to unbounded
- Cluster Coordination
- Restrict testing of legacy discovery to tests
- EQL
- Return sequence join keys in the original type
- Features/Data streams
- "no such index [null]" when indexing into data stream with
op_type=index [ISSUE]
- Data streams: throw ResourceAlreadyExists exception
- Track backing indices in data streams stats from cluster state
- Features/ILM+SLM
- Fix race in SLM master/cluster state listeners
- Features/Ingest
- Fix handling of final pipelines when destination is changed
- Fix wrong pipeline name in debug log
- Fix wrong result when executing bulk requests with and without
pipeline
- Update regex file for es user agent node processor
- IdentityProvider
- Only call listener once (SP template registration)
- Machine Learning
- Always write prediction_probability and prediction_score for
classification inference
- Get data frame analytics stats API can return multiple responses if
more than one error
- Ensure .ml-config index is updated before clearing anomaly job’s
finished_time
- Ensure annotations index mappings are up to date
- Handle node closed exception in ML result processing
- Recover data frame extraction search from latest sort key
- SQL
- Fix NPE on ambiguous GROUP BY
- Fix SYS COLUMNS schema in ODBC mode
- Search
- Disable sort optimization on search collapsing
- Search fix: query_string regex searches not working on wildcard
fields
- Snapshot/Restore
- Cleanly Handle S3 SDK Exceptions in Request Counting
- Fix Concurrent Snapshot Create+Delete + Delete Index
* Upgrades:
- Infra/Core
- Upgrade to Lucene 8.6.2
PR: 249106
Submitted by: Juraj Lutter <juraj@lutter.sk>
This will result in linker errors once D26044 is commited since
currently duplicates are silently filtered out.
While here, add LICENSE variables.
Submitted by: arichardson
Reviewed by: zeising
Differential Revision: https://reviews.freebsd.org/D26371
