JProfiler 16: baixar versions anteriors

New features:
- MCP server: Heap dumps can be taken in live sessions and existing snapshots can be analyzed with class histograms, biggest objects, instance queries with value filters, reference chains, paths to GC roots, dominator chains and automated heap inspections. Heap walker scripts can be used for filtering and custom grouping.
- MCP server: Individual probe events with back traces, event histograms and event timelines can be queried. Open control objects, for example unclosed JDBC connections, are listed for leak analysis together with the stack trace of the open call.
- MCP server: GC data is now available, including individual garbage collections with causes, pause times, phases and heap usage, as well as the GC configuration and flags
- MCP server: Monitor analysis is now available, including monitor statistics, contention timelines, locking history with stack traces and lock graphs. Monitor dumps can be taken from live JVMs for deadlock analysis.
- MCP server: Thread dumps are now available, including virtual thread hierarchies
- MCP server: MBean snapshots are now available
- MCP server: All telemetries are now available
- MCP server: Allocation recordings is now available
- MCP server: Method complexity analysis with curve fits is now available
- MCP server: JavaScript origin tracking now provides browser call stacks to the agent
- MCP server: CPU analysis now supports thread and thread group selection, thread status filtering, per-thread attribution of hot spots and call tree nodes, and outlier detection with exceptional method recording
- MCP server: Triggers with actions for thread dumps, heap dumps and MBean snapshots can be configured in the recording configuration
- MCP server: Script probes and custom probes can now be recorded and analyzed
- MCP server: Heap dumps, monitor dumps, thread dumps, MBean snapshots and recording control are available as live actions on running JVMs
- Allocation recording now uses byte-interval sampling with the low-overhead JVMTI allocation sampling mechanism (JEP 331) as the default on Java 11+, reducing recording overhead and avoiding interference with escape analysis
- Added a Maven plugin with goals for profiling and for exporting, comparing and analyzing snapshots, matching the existing Gradle plugin
- The Gradle tasks and Maven goals can now automatically download a JProfiler installation if none is configured
- AI probe: Added support for LangChain4j async and streaming requests
- The biggest objects view in the heap walker now has "Group by container" grouping for framework containers such as Spring application contexts, servlet sessions, servlet contexts and Hibernate persistence contexts
- Added support for MongoDB 5. 13
- Kafka probe: Added support for share consumers and Streams polling
- Added support for the MariaDB JDBC driver 4. 0
- Added support for the WildFly 42 EJB client
- Cassandra probe: The consistency level is now recorded as an additional event field
- JFR snapshots: Wall-clock recordings from async-profiler are now supported, samples from sleeping threads are shown under the "Waiting" thread status
- Process pipe reads, waits for process exits and blocking console reads are now classified as NET I/ O.
- Thread history view: Added a copy to clipboard action for thread rows
Bug fixes:
- GC summary statistics could be permanently wrong when GC data arrived faster than the UI could process it
- Probe telemetry views could show the data of another telemetry under the wrong name in snapshots
- Quick search in the JFR hot spots view and exporting an entire hot spots tree could hang indefinitely
- Fixed a deadlock between quick search and tree node expansion
- Hot spots views with invocation counts showed wrong percentages after hiding nodes
- Heap walker: "Less than or equal" filters on float and double fields wrongly excluded zero and negative values
- HTML export of locking graphs showed the waiting thread as the owning thread
- Quotes in CSV export value

Bug fixes:
- IDE integrations did not work on macOS if the JProfiler MCP server was running
- IDE integrations: The connection between JProfiler and the IDE could be lost, for example, when macOS App Nap was active
- VS Code integration: Fixed window activation with current Visual Studio Code versions on macOS
- VS Code integration: Fixed profiling launches with string VM arguments and debug session cancellation during startup
- VS Code integration: Fixed various source navigation problems, including lambdas compiled with Java 24+, generic types, multi-root workspaces and multi-release JARs
- VS Code integration: Automatic JProfiler discovery through the PATH environment variable did not work on Linux
- Fixed source navigation for proxy classes
- AI probe: langchain4j embedding models that implement the EmbeddingModel interface with default methods, such as the OpenAiEmbeddingModel, were not intercepted
- CPU profiling did not work for threads that were started before the VM initialization phase, for example, by other native agents
- On Java 27, the internal "Master-Poller" system threads were not ignored in the CPU views
- Fixed a race condition in the profiling agent when recording monitors
- An error in a telemetry, for example, in a JMX telemetry, could stop all further telemetry recording
- GC probe: The GC-specific flags for the parallel garbage collector were not shown correctly in the "GC flags" view
- JavaScript origin tracking: The connection between the "JProfiler Origin Tracker" Chrome extension and the JProfiler UI could be lost after a period of inactivity

New features:
- Offline profiling can now be configured directly in the -agentpath VM parameter with common profiling settings and a complete automated recording sequence, without requiring a JProfiler config file, the controller API or jpcontroller
- jpenable can now configure offline profiling with common profiling settings and a complete automated recording sequence, interactively or with command line options, without requiring a JProfiler config file
- The Gradle and Ant profiling tasks can now configure offline profiling with common profiling settings and a complete automated recording sequence directly with task attributes, without requiring a JProfiler config file
- MCP server: Made the profiling workflow more robust for weaker models and added support for the Antigravity AI coding tool
- jpexport: Added a -maxEvents parameter for the ProbeEvents export to limit the maximum exported file size for huge event counts. The default is 100000
- JPA/ Hibernate probe: Added support for Hibernate stateless sessions and Hibernate Data Repositories
- Improved stack trace hints in the monitor views and for the thread creation stack trace
- The embedded agent controller API can now be shaded into your application with the -Djprofiler. controllerPackage= [renamed package] system property
Bug fixes:
- JProfiler executables did not exit if the native profiling library could not be found
- Fixed a memory leak in the frontend when recording probe events
- Fixed a race condition when exporting probe events
- Fixed a startup race condition in the MCP server

New features:
- Added an MCP (Model Context Protocol) server so AI coding agents can interact with JProfiler. Set up your agent from the JProfiler UI or with the installation script available on our website.
- All command-line executables now feature a modern rich console with arrow-key menus, inline completion, progress indicators and styled output
- Linux profiling agents are now included in all distributions, so they do not have to be auto-downloaded at runtime when profiling containerized JVMs or JVMs on remote Linux servers
- Support new lambda numbering scheme (Java 24+) for precise source navigation in IntelliJ IDEA
Bug fixes:
- When setting the zoom factor for the UI, the factor was applied twice
- Heap walker: The biggest objects calculation could not be canceled except for the initial object set

New features:
- An AI probe was added
- New UI for the IDE and updated artwork for installers
- Streamlined installer experience without requiring admin privileges by default
- UI scaling and support for running the JProfiler UI on Linux with all OpenJDK variants
- Added support for huge numbers of recorded events and control objects and removed consolidation of probe events. In the advanced probe session settings, you can configure the maximum file size for probe events.
- Added a secondary timeline histograms view for all filtered events. An event filter can be added by dragging in this view.
- Added a secondary view to show CPU hot spots, probe hot spots and the probe call tree for all filtered events
- Added secondary histogram views for duration and for throughput (where applicable)​. An event filter can be added by dragging in these views.
- Added support for horizontal scrolling in the probe event views when the viewport width falls below the total minimum column width
- Throughput metrics in Socket and Files Probes
- Added column totals for all events in the JFR event browser
- Data in the secondary views of the JFR event browser is available immediately
- When sorting events in the JFR browser, all events are sorted, not just the limited cutoff set as before
- Tables with a cutoff now show the number of displayed and the total number of rows together with a hyperlink at the bottom to add more rows
- The selection panel in the event browser now shows the number of selected events
- The new jdk.​CPUTimeSample event (Java 25+) is now used instead of the jdk.​ExecutionSample event for calculating data in the CPU section if it has been recorded.
- Added all experimental JFR events to the JFR browser
- Added support for throttle settings in the JFR recording wizard. These settings are now also shown in the thread status help tooltip.
- Added support for virtual thread dumps
- Added support for grouping similar threads in thread dumps
- Percentages for threads in the find threads dialog
- Multiple selection in hot spot views with cumulated statistics in the node detail dialog
- Columns with long text now show an expansion button when hovering over cells to open the node detail dialog with the full text. Double-clicking on such cells has the same effect.
- Heap walker reference views: The node detail dialog now shows the full strings for string objects and char arrays
- Added support for deobfuscating snapshot comparisons
- Added an action to pin rows in timeline views at the top
- Added a new sort drop-down in timeline views with tooltips and a sort direction chooser
- Default sort directions in timelines have been changed to show newer long-lived live objects or more recently closed rows at the top
- Added full support for profiling Java 26
- Added a way to clear the current selection in the call tree analysis drop-down menu to apply the analysis to the entire tree
- Highlight edges in graph views when hovering over a node
- Changed auto-updated tables to keep the selection in the same position relative to the viewport to avoid loss of context
- Added support for compiling scripts with target Java 25
- HTTP client probe: Added support for HttpClient with HTTP/​3
- Improved robustness of snapshot saving in low memory conditions
- Added support for a very large number of GC roots that are possible with virtual threads on Java 25+
- Added a -jvm=​/path/​to/​jreBundle VM parameter to set the used JRE explicitly. This is supported in the jprofiler.​vmoptions file and with -J-jvm=​... as a command line argument.
Bug fixes:
- A heap dump with only recorded objects that was triggered via the API or a trigger in a live profiling session was not handled correctly in the UI
- Fixed heap dumps with virtual threads on IBM JVMs