BreakTest 2026.07.11

BreakTest 2026.07.11 is a test-plan portability, parallel-runtime stability,
and AI Repair reliability release built from the changes since BreakTest
2026.07.10. It introduces compressed JMX archives that can carry recorded
request and response evidence, fixes concurrency hazards in parallel and forked
execution, and makes automated GUI repair safer and recoverable.

Highlights

• Saves .jmx plans as compressed ZIP archives while continuing to load
  legacy XML JMX files.
• Embeds filtered HAR or replay recordings with the test plan so recorded
  evidence remains available after the plan is moved or reopened.
• Adds recording-retention choices for all traffic, omitted static bodies,
  omitted static resources, or no embedded recording.
• Adds Store Replay to View Results so the latest replayed requests and
  responses can become the plan's current recording evidence.
• Isolates mutable sampler and engine state between parallel ForEach and Fork
  workers, eliminating timing-dependent configuration and status races.
• Fixes GUI update checks so they run immediately at startup, repeat every six
  hours by default, and tolerate unrelated release assets.
• Hardens AI Repair with conflict-aware native Regex Extractor planning,
  atomic rollback, safe live-tree deletion, backup restoration, and explicit
  blocked completion reporting.

Compressed JMX Archives And Recordings

• Newly saved JMX plans contain the XML test plan and referenced attachments
  in one compressed archive, while retaining the .jmx filename extension.
• Loading detects ZIP-backed plans automatically and retains support for
  existing plain-XML JMX files.
• Imported HAR recordings are filtered to the generated samplers and stored in
  a native manifest with request and response bodies externalized as deduplicated
  archive entries.
• Recording references and checksums are retained on HTTP samplers, including
  after GUI edits and save/load round trips.
• Archive entry names, linked content, and required recording dependencies are
  validated without extracting files to the filesystem.
• Test validation no longer rewrites the open archive, and GUI saves show an
  activity overlay while the archive is assembled.

Replay And Recorded Evidence

• View Results can store the latest replayed sample for every resolved sampler
  as a replacement recording bundle.
• Replay storage supports the same four retention levels as HAR import.
• Recorded request and response data remains available to Sampler Result and
  recorded-HAR comparison views after reopening a plan.
• Sampler lookup now resolves nodes consistently across ordinary controllers
  and Test Fragments when replay results are associated with the plan tree.

Parallel And Fork Runtime Stability

• Parallel ForEach branches now clone child samplers as well as controllers,
  preventing siblings from racing while merged configuration is applied and
  recovered.
• Compiled sample packages resolve cloned samplers back to their source across
  nested parallel sections, including clone-of-clone cases.
• Worker-local variables isolate LAST_SAMPLE_OK and transaction package state
  while other virtual-user variables remain shared for intentional branch
  communication.
• The parent worker now receives a deterministic all-branches success result
  instead of whichever branch wrote its status last.
• Fork workers no longer overwrite the main virtual user's transaction state
  or notify the wrong transaction listeners.
• Parent-mode Transaction Controllers in parallel or forked branches retain a
  source-controller chain, preventing swallowed package-reset errors when
  sibling or nested clones interleave.

GUI Update Reliability

• Automatic update checks now run immediately at GUI startup and then repeat at
  breaktest.update.interval_hours, whose default is now six hours.
• Persisted check timestamps only deduplicate concurrently running instances
  instead of suppressing expected periodic checks.
• Numeric qualifier segments compare correctly, so for example rc10 sorts
  after rc2.
• Release parsing strictly validates the expected binary ZIP and checksum but
  ignores unrelated or malformed extra assets.
• Corrected the documented settings to breaktest.update.enabled and
  breaktest.update.interval_hours.

AI Repair Safety And Recovery

• Repair planning deduplicates conflicting dynamic literals, including encoded
  variants, and emits native Regex Extractor actions only when the recorded
  evidence supports a safe pattern.
• Regex planning now handles quoted token values without producing unsupported
  extractor forms or duplicate replacements.
• Applying a repair batch is transactional: failed actions are rolled back
  through GUI undo, with the pre-run backup used when tree integrity cannot be
  preserved.
• Added restore_open_plan_from_backup for explicit recovery after a reported
  live-tree integrity failure.
• Live node deletion now validates exact membership and match counts, rejects
  protected container deletion, and suppresses unsafe selection side effects.
• Stale node IDs fall back to stable paths after undo or reload, avoiding edits
  against invalid GUI tree references.
• GUI refresh is null-safe, bridge failures retain full stack diagnostics, and
  incomplete repair runs must report Status: blocked rather than appearing
  successful.

Tests

• Added archive save/load, attachment, recording-filter, replay, sampler
  resolution, and GUI lifecycle regression coverage.
• Added parallel sampler-isolation, worker-local status, Fork transaction
  state, nested clone resolution, and parent-transaction reset coverage.
• Added update scheduling, qualifier comparison, and extra-asset parsing
  coverage.
• Added AI Repair regression tests for conflicting literals, quoted-token
  regexes, rollback status, safe deletion, orphan extractors, stale tree state,
  and null-safe refresh.

Compatibility Notes

• Java 21 or later is still required.
• Legacy plain-XML JMX plans continue to load, but newly saved compressed JMX
  archives require a BreakTest version that understands the archive format.
• Existing JMeter-compatible test-plan elements remain supported where
  practical; embedded recordings are a BreakTest extension.
• This release uses the direct Git tag 2026.07.11.

BreakTest 2026.07.10

BreakTest 2026.07.10 is a workflow, visualization, and scalability release
built from the changes since BreakTest 2026.07.08. It adds direct HAR import,
parallel ForEach execution, GUI-assisted updates, a richer View Results
listener, and several targeted runtime and desktop performance improvements.

Highlights

• Added a guided HAR import wizard for uncompressed .har recordings, with
  hostname filtering, request grouping, delay options, and automatic test-plan
  generation.
• Added optional parallel execution to the ForEach Controller, with a
  configurable maximum concurrency and lazy branch creation for very large
  collections.
• Expanded View Results with a table layout, selectable columns, response-time
  data, request/response sizes, URLs, filtering, and improved status icons.
• Reduced repeated test-tree traversal by caching parent-controller paths used
  by loop control actions.
• Added the HTTP request URL to the Sampler Result text view and made the
  initial test-plan pane width configurable.
• Added GUI-only automatic and manual checks for stable BreakTest releases,
  with verified download, post-exit installation, rollback, and restart.

HAR Import

• Added File > Import HAR... and a matching HAR toolbar action.
• Added guided steps for selecting a HAR file, choosing hostnames, and
  configuring import behavior.
• Added support for ignoring recorded error responses, continuing on error,
  adding request indexes, detecting dynamic URLs, and splitting transactions
  using recorded idle gaps.
• Added recorded, fixed, random, Gaussian, and disabled transaction-delay
  modes.
• Generated plans can include Transaction Controllers, Parallel Controllers
  for concurrent requests, HTTP Request Defaults, Cookie Managers, assertions,
  shared headers, and request-local headers.
• Preserved BreakTest HAR metadata so imported requests remain compatible with
  recorded-HAR inspection and repair workflows.
• Fixed reference and variable handling in imported requests.

Parallel ForEach And Runtime Performance

• Added Parallel execution and Maximum parallel executions options to
  the ForEach Controller.
• Branches are created only when a concurrency slot becomes available, so a
  million input values no longer require a million controller clones up front.
• Transaction-controller source mappings are scoped to each active branch and
  released with that branch, keeping retained execution state proportional to
  configured parallelism rather than total item count.
• Preserved branch-local ForEach output values while sharing the surrounding
  virtual-user context and honoring existing same-user iteration behavior.
• Added parent-path caching for loop actions that previously traversed the test
  tree repeatedly after errors or logical loop transitions.

Results And Desktop Experience

• Added Tree and Table modes to View Results, with table results above the
  detail pane and the existing tree/detail split retained in Tree mode.
• Added timestamp, time, latency, connect time, request size, response size,
  thread group, thread name, label, and URL columns with configurable
  visibility.
• Added filtering by thread group, thread name, and sampler label, including a
  table context-menu action for filtering to a selected thread.
• Fixed automatic scrolling and filter controls during live table refreshes.
• Added full-value tooltips for truncated table cells and numeric sorting for
  timing and size columns.
• Modernized success/failure icons and added HTTP URLs to the Sampler Result
  text tab.
• Set a stable startup width for the test-plan pane, configurable through
  jmeter.gui.testplan.width.

GUI Updates

• Added automatic stable-release checks in GUI mode, limited to once every 24
  hours by default, plus a manual Help > Check for Updates action.
• Added update notifications with links to the matching GitHub release notes.
• Downloads the expected binary ZIP and verifies its published SHA-512
  checksum and GitHub SHA-256 asset digest before installation.
• Installs through a post-exit helper so active JAR files are not replaced,
  rolls back failed replacements, and restarts BreakTest after success.
• Preserves user and system properties, third-party lib/ext plugins, and
  user-installed driver JARs during updates.
• Refuses source checkouts and malformed distributions, and confines archive
  extraction and installation paths to prevent traversal or symlink escapes.
• Update checks are disabled outside the GUI and can be configured through
  breaktest.update.check.enabled and breaktest.update.check.interval_hours.

Benchmarks And Tests

• Added high-capacity ForEach benchmarks covering active-window memory,
  representative sample payloads, and parent-mode Transaction Controllers.
• Added regression tests for lazy branch creation, bounded transaction mapping,
  branch-local variables, parent transactions, table sorting, filtering,
  automatic scrolling, URL display, and visualizer layout behavior.
• Expanded HAR conversion tests across grouping, headers, request bodies,
  delay modes, hostname filtering, and generated configuration elements.

Compatibility Notes

• Java 21 or later is still required.
• Existing JMeter-compatible JMX files remain supported where practical.
• Parallel ForEach execution is opt-in; existing ForEach Controllers remain
  sequential unless explicitly enabled.
• Automatic update checks run only in the GUI; command-line and non-GUI test
  execution do not contact the update service.
• This release uses the direct Git tag 2026.07.10 rather than a rel/...
  prefix.

BreakTest 2026.07.08

BreakTest 2026.07.08 is a focused stability and GUI-assistance release built
from the changes since BreakTest 2026.07.07. It keeps the same Java 21+
baseline while improving fork/parallel execution, AI-assisted GUI repair
actions, and day-to-day desktop ergonomics.

Highlights

• Fixed Fork Controller lifecycle cleanup so completed forks are removed while
  the parent virtual user keeps running.
• Added ctx.getThread().stopForks() and ctx.getThread().stopForksNow() for
  Groovy/JSR223 scripts that need to end active fork branches from the main
  thread.
• Improved fork shutdown so timers, pacing waits, active fork samplers, and
  virtual-thread fork workers are woken promptly.
• Fixed fork branch behavior for "Start Next Thread Loop on error" so failed
  fork samples stop the fork branch locally instead of continuing to later
  requests.
• Hardened Parallel Controller execution so nested loop branches keep their own
  loop indexes and sampler package configuration is synchronized safely across
  concurrent branches.
• Added a GUI-backed clone_node_open_plan MCP tool so AI-assisted repair can
  copy full controller/sampler subtrees, including Transaction Controller
  pacing settings, without falling back to raw JMX reconstruction.
• Improved AI Helper and log-panel usability with Shift+Enter newlines in the
  JSR223 AI Helper prompt and a visible bottom-log resize boundary.
• Reduced noisy delay-interruption warnings during intentional fork/test stops.

Fork And Parallel Controller Stability

• Completed fork tasks, executors, and active-controller mappings are now
  cleaned up promptly instead of waiting for thread teardown.
• Fork workers now track current timers, samplers, and worker threads so
  graceful and forced fork stops can wake the correct in-flight operation.
• stopForks() requests graceful fork completion and wakes fork timers without
  stopping the main virtual user.
• stopForksNow() cancels active fork tasks, shuts down fork executors, and
  stops fork samplers that implement StoppableSampler.
• Intentional fork-stop interruptions no longer emit WARN-level "Loss of delay"
  timer messages.
• Fork timer/sampler bookkeeping now removes active entries by identity so
  equally configured sibling branches cannot remove the wrong entry.
• Fork branches now consume local logical error actions and reset
  LAST_SAMPLE_OK so fork control-flow decisions do not leak into the main
  virtual-user path.
• Parallel branch sampler configuration is protected from concurrent compiler
  state corruption.

AI-Assisted GUI Repair

• Added clone_node_open_plan to the BreakTest MCP bridge for cloning a node
  and its complete child subtree in the live GUI plan.
• The clone tool supports stable nodeId selectors from find_open_plan_nodes
  and path/index fallbacks, matching the existing move/delete structural tools.
• The clone operation uses the same full subtree copy path as the GUI Duplicate
  action, preserving element properties and children rather than rebuilding
  controllers from partial fields.

Desktop Polish

• The docked Logs / AI Auto Scripting panel now has a visible top divider and a
  wider split-pane drag target, making the resize area discoverable without
  hovering.
• The JSR223 AI Helper request text area now treats Shift+Enter as an explicit
  newline.
• Extractor and post-processor icons in the test-plan tree were aligned more
  closely with the Add menu iconography.

Benchmarks And Tests

• Added a JMH benchmark for Fork Controller capacity, including lightweight and
  more representative fork branch scenarios.
• Added regression tests for fork cleanup, fork stopping, fork error handling,
  nested parallel loop indexes, and tree icon rendering.

Compatibility Notes

• Java 21 or later is still required.
• Existing JMeter-compatible JMX files remain supported where practical.
• stopForks() and stopForksNow() are BreakTest extensions intended for
  scripts that explicitly need to manage asynchronous fork branches.

BreakTest 2026.07.07

BreakTest 2026.07.07 is a focused desktop and reporting update built from the
changes since BreakTest 2026.07.03. It keeps the same Java 21+ baseline and
JMeter-compatible test-plan model while improving the day-to-day GUI workflow,
local settings management, validation, and live performance reporting.

Highlights

• Added a configurable Performance Report listener with percentile, throughput,
  bandwidth, error, and connect-time columns.
• Added a Settings dialog backed by the full JMeter properties catalog, with
  search, default/modified indicators, reset controls, and safe persistence to
  local override property files.
• Modernized the BreakTest desktop UI with a refreshed application icon,
  modern toolbar icons, cleaner tree rendering, improved menu organization, and
  better light/dark FlatLaf styling.
• Improved View Results Tree and validation ergonomics with detach/dock support,
  validation targeting, auto-detach controls, collapsible search/file panels,
  and comment visibility controls.
• Moved sampler-scoped HTTP headers into the HTTP Request sampler with
  automatic migration from child HTTP Header Manager elements.
• Fixed saved test plans with the new Performance Report listener so they
  reload correctly.
• Added child-count badges in the test-plan tree and compact Transaction
  Controller delay summaries beside transaction names.
• Added counts to HTTP Request Params, Headers, Body, and Files tabs so hidden
  request data is visible at a glance.

Performance Reporting

• Added a dedicated Performance Report visualizer for configurable live
  aggregate metrics.
• Added optional percentile columns including median, P75, P90, P95, and P99.
• Added controls to ignore error response times in response-time metrics.
• Added resource-use guidance for percentile columns, since they retain richer
  response-time distribution data while enabled.
• Added tests for column configuration, percentile tracking, error handling, and
  report reset behavior.

Settings Management

• Added a Settings page available from the GUI menu and toolbar.
• Added typed setting definitions, editors, grouping, search, default-value
  display, modified badges, and reset/save actions.
• Added a full settings catalog resource for JMeter/BreakTest properties.
• Persisted changes to local user/system override properties rather than
  modifying the bundled jmeter.properties file.
• Added tests for catalog loading, property storage, and model behavior.

Desktop UI And Validation

• Refreshed the main frame, toolbar, menu factory, tree renderer, logger panel,
  target-rate chart, and scripting log presentation.
• Added modern SVG toolbar assets and updated icon mappings for common actions.
• Improved validation controls so selected thread groups can be targeted more
  directly.
• Added View Results Tree detach/dock workflows and clearer validation result
  viewing.
• Added native HTTP Request headers so sampler-specific headers live on the
  request itself, while higher-level Header Managers still apply by scope.
• Updated the Transaction Controller icon so it reads as a transaction span
  rather than a small generic controller/action glyph.
• Added direct-child count badges to tree nodes and compact delay summaries for
  Transaction Controllers with built-in delay.
• Added value counts to HTTP Request data tabs for Params, Headers, Body, and
  Files.
• Added tests around tree-listener behavior.

Compatibility Notes

• Java 21 or later is still required.
• Existing JMeter-compatible JMX files remain supported where practical.
• Percentile columns in the Performance Report intentionally keep additional
  response-time history only while those columns are selected.
• Most Settings changes are written as local overrides and take effect after
  restarting BreakTest.

Security Notes

BreakTest continues to use the trusted-test-plan model. Treat JMX files as
executable input and run untrusted plans only in an isolated environment. See
SECURITY.md and THREAT_MODEL.md.

BreakTest 2026.07.03

BreakTest 2026.07.03 Release Notes

BreakTest 2026.07.03 is the first public BreakTest release: an independent,
Apache-2.0, JMeter-compatible continuation focused on a leaner runtime, modern
HTTP, better debugging, and a smoother path from local scripts to scaled
execution.

BreakTest keeps the useful JMeter test-plan model and JMX workflow, while
moving the engine toward Java 21+, HTTP/2, lower resource usage, visual
diagnostics, AI-assisted repair workflows, and BreakTest Enterprise scale.

Highlights

• Modern HTTP path based on Apache HttpClient 5 with first-class HTTP/2 support.
• Lower memory and CPU pressure through lazy decompression, retention modes,
  lightweight cloning, lazy diagnostics, and allocation reductions.
• Java 21+ baseline with virtual-thread support.
• Parallel Controller and Fork Controller for realistic concurrency and
  independent branches of work.
• Unified Thread Group scheduling with closed/open workload models, phases,
  open-model helpers, pacing, and pause/resume support.
• Undo/redo support for common GUI test-plan edits.
• Background JMX loading and safe missing-plugin placeholders.
• Richer View Results Tree diagnostics, source navigation, and HAR-backed
  recorded-vs-replayed request/response diff views.
• Easier If/While conditions, transaction timing controls, random CSV ordering,
  CSV preview, and extractor fail-on-no-match options.
• Apache License 2.0, with Apache JMeter attribution retained.

Modern HTTP Runtime

• Added Apache HttpClient 5 sampler support for HTTP/1.1.
• Added an async HttpClient 5 HTTP/2 sampler path with HTTP/2 and HTTP/2
  preferred protocol selection.
• Added negotiated fallback, connect-time reporting, HTTP/2 upload content
  length handling, and safer HTTP/2 replay behavior.
• Reduced HTTP/2 resource usage by limiting reactor threads by default and
  avoiding unnecessary async client/reactor rebuilds between same-user
  iterations.
• Added conservative HTTP/2 closed-session retry handling for common HC5
  connection-closure failures.
• Fixed NTLM authentication on the HttpClient 5 path, including variable
  resolution before async auth callbacks.
• Kept HttpClient 4 runtime jars available for legacy plugin compatibility while
  keeping the built-in HTTP Request sampler on HttpClient 5.
• Added Brotli and Zstandard response decoding alongside gzip and deflate.

Resource Usage Improvements

• Lazy response decompression stores compressed bytes and decompresses only when
  response data is accessed.
• Response-processing modes can store compressed data, discard successful
  bodies, keep only failing bodies, or retain MD5 checksums.
• Lightweight test-plan cloning shares safe property state where possible and
  reduces per-thread object churn.
• Header merging, temporary property handling, and response read buffers were
  optimized for hot-path sampler execution.
• Rich SampleResult metadata is now lazy and listener-driven, avoiding
  unnecessary per-sample storage when diagnostics are not needed.
• A throughput regression in temporary property recovery was fixed while keeping
  the correctness improvements for identity semantics.

Controllers, Scheduling, And Flow Control

• Added native Parallel Controller execution with bounded parallel sampler
  scheduling and HTTP state safeguards.
• Added Fork Controller for asynchronous child branches that share virtual-user
  context and complete before thread teardown.
• Added unified Thread Group scheduling controls with closed/open model
  selection.
• Added open-model schedule helpers, even-arrival scheduling, max active thread
  limits, safe schedule parsing, and graph preview support.
• Added closed-model phases so a schedule can combine different ramp-up speeds,
  hold periods, and thread targets. This makes it possible to shape almost any
  closed workload model from one Thread Group schedule.
• Added Thread Group pacing controls.
• Cleared runtime user variables on new iterations when same-user mode is
  disabled, while preserving initial variables.
• Added Pause and Resume in the GUI and on the existing UDP command port, with
  bin/pause.* and bin/resume.* wrappers. Pausing is useful during ramp-up
  when you want to stop increasing load and hold the current point. In closed
  model, active threads stay active. In open model, new arrivals are paused and
  active threads finish naturally, so load drops while paused.
• Improved shutdown so it is quicker: shutdown no longer waits for timers to
  complete and only waits for in-flight server responses to finish cleanly.

GUI And Test-Plan Editing

• Added undo and redo for semantic tree edits including add, delete, update,
  move, and search/replace operations.
• Added background JMX loading with a loading overlay.
• Added opt-in fast GUI loading through jmeter.gui.load.fast=true.
• JMX files with missing plugin elements now load with disabled placeholders
  rather than failing the whole file.
• Added placeholder GUI panels and clear notifications for missing plugin
  elements.
• Improved Search Tree flagging for pre-processors, post-processors,
  assertions, timers, and config elements.
• Added selective "Remove Matching" for search results with confirmation so bulk
  cleanup can avoid deleting parent nodes accidentally.
• Improved GUI layout and editor resizing around controller panels and tree
  editors.

Scripting, Data, And Correlation

• Added structured If and While Controller condition rows with all/any matching,
  while preserving legacy condition expressions.
• Added loop/while index behavior options and exported index variable display.
• Added Transaction Controller measurement modes plus built-in delay and pacing
  controls.
• Added extractor fail-on-no-match options for Boundary, CSS/HTML, Regex,
  XPath, XPath2, JSONPath, and JMESPath extractors.
• Added random-order reading for CSV Data Set records.
• Added CSV preview of the first sample variable assignments.
• Focused scripting on JSR223/Groovy by removing legacy BeanShell, BSF/JEXL2,
  Rhino JavaScript, and LogKit paths.

Visual Debugging And HAR Migration

• View Results Tree now shows richer request/response diagnostics, endpoint
  details, HTTP/TLS metadata, cookies, variables, binary/text detection, and
  jump-to-source.
• Rendered response views and raw text response views now keep the expected
  response display lifecycle.
• BreakTest JMX metadata preserves origin and source-path information for
  debugging tools.
• HAR-backed recorded request/response tabs appear in View Results Tree and HTTP
  sampler editors when matching BreakTest HAR metadata is available.
• Recorded-vs-replayed diffs make HAR migration and HTTP/2 replay debugging
  easier.
• Transaction child samples preserve source test-element paths so folded
  transaction results can still navigate to the correct source element.
• HTTP/2 replay removes unsupported hop-by-hop headers and normalizes
  request/status lines for cleaner diffs.
• Host-only cookie matching is stricter so cookies do not leak to unrelated
  hosts or subdomains.

Compatibility Notes

• Existing JMeter JMX files remain supported where practical.
• Many package names, property names, command internals, and Maven coordinates
  intentionally still use jmeter or org.apache.jmeter for compatibility.
• BreakTest-specific JMX metadata may not round-trip into older JMeter versions.
• Plugins that depend on removed internals may need migration.
• Legacy HttpClient 4 jars are bundled for plugin compatibility, but the built-in
  HTTP Request sampler runs on HttpClient 5.
• RMI Remote Server style distributed execution is removed. Use BreakTest
  Enterprise or another controlled orchestration layer for distributed tests.
• Java 21 or later is required.

Security Notes

BreakTest inherits Apache JMeter's trusted-test-plan security model. Treat JMX
files as executable input: they can run scripts, load classes, read files, make
network requests, and interact with systems reachable from the runner.

Only open or run test plans you trust, or isolate them first. See
SECURITY.md and THREAT_MODEL.md for the
project security model and vulnerability reporting process.

Legal And Attribution

BreakTest is licensed under the Apache License, Version 2.0. BreakTest is a
derivative work of Apache JMeter, and Apache JMeter copyright and attribution
notices are retained in the source tree and distribution notices.

BreakTest is independent from The Apache Software Foundation. It is not
affiliated with, endorsed by, or sponsored by The Apache Software Foundation.