Release notes
These release notes describe the relevant changes implemented for each SonarQube Server release version since the 2026.1 LTA version.
These release notes describe the relevant changes implemented for SonarQube Server 2026.4. If you’re upgrading from the previous LTA, see the 2026.1 LTA to LTA release notes. For a complete list of all changes, see the Full release notes.
New and enhanced features
View the release notes for new and enhanced features for SonarQube Server.
2026.4
AI capabilities
Sonar way for agentic AI quality gate
A new built-in quality gate, Sonar way for agentic AI, provides quality standards tuned for agentic development workflows. Its conditions cover reliability, security, and maintainability severity on new code, plus new coverage, new duplication, and a dependency-risk condition, so high-risk changes are caught before they merge. The gate adapts its conditions automatically to the instance's active mode (Standard experience or MQR). The previous built-in Sonar way for AI Code becomes a custom quality gate for instances that were using it, keeping its existing conditions and project assignments intact. For more information, see Quality gate for agentic AI.
Quality gate history
A new Quality gate history page shows how a project's quality gate verdict has evolved across released versions on the main branch. Each released version is marked safe (gate passed) or risky (gate failed) at release time, with a summary of safe versus risky releases and period filtering. The page helps teams monitor guard-rail adherence over time and spot high-risk releases. It requires the Previous version new code definition and is reachable by going to your project > Quality gate history. For more information, see Quality gate history.
Automated GitHub App creation
Connecting SonarQube Server to GitHub no longer requires manually creating a GitHub App and copying your credentials by hand. From Administration > Configuration > General Settings > DevOps Platform Integrations, administrators can launch a guided flow that uses a GitHub App Manifest to pre-configure the exact permissions, webhook, and callback URLs SonarQube needs. The feature automatically retrieves and saves the resulting App ID, client ID, client secret, and private key. The flow works with both GitHub.com and GitHub Enterprise Server, and falls back to manual setup when the automated handshake is not possible. For more information, see Setting up a GitHub App.
Severity-based quality gate conditions for new code
Quality gates can now use severity-based conditions on new code. New metrics expose the highest severity among new issues for each issue type in the Standard experience and for each software quality in MQR mode, so you can fail a quality gate when new code introduces an issue at or above a chosen severity. The metrics update live as issue severities change, without requiring a full re-analysis. For more information, see Managing custom quality gates.
Consistent new code fallback for reference branch analysis
When the Reference branch new code definition is used and SCM data is missing or invalid (for example, on a shallow clone or a repository without Git history), new code is now computed consistently across the first and subsequent analyses. SonarQube Server compares the current analysis against the last snapshot of the configured reference branch, falling back to the last snapshot of the main branch when the reference branch is not available. This removes the fluctuating new-code calculations that could previously appear between runs. For more information, see Configuring new code calculation.
Faster pull request analysis for large Java and C# projects
Incremental analysis in the taint engine significantly reduces pull request analysis time on large, complex Java and C# projects, where scans previously took the longest. Most projects already finish quickly and won't notice a difference, but on the slowest-scanning projects, pull request analysis is considerably faster.
New MISRA C compliance quality profiles
Check your C code against the MISRA C guidelines with the new Sonar MISRA compliance quality profiles. Available in Enterprise edition and higher, each profile combines the Sonar way quality profile with rules mapped one-to-one to the matching MISRA C guidelines, giving you both Sonar and MISRA expertise in a single profile. The following profiles are now available:
Sonar MISRA C:2012 Compliance
Sonar MISRA C:2012 AMD1 + TC1 Compliance
Sonar MISRA C:2012 AMD2 Compliance
Sonar MISRA C:2012 AMD3 Compliance
Sonar MISRA C:2023 Compliance
New rule tags (misra-c-2012, misra-c-amd1-tc1, misra-c-amd2, misra-c-amd3, and misra-c-2023) let you identify the corresponding rules and build custom MISRA C profiles from scratch. See Understanding quality profiles for more information.
External issue import with overlapping start and end locations
The generic issue import format now accepts findings whose start column equals their end column. Such issues are imported and displayed correctly, expanding the highlight to the end of the line, so external reports that point to a single position in the code no longer fail to import. For more information, see Generic formatted reports.
Software architecture analysis
The software architecture capability is now available in SonarQube Server. Architecture analysis runs on each scanner execution with no additional configuration and is included in all commercial editions (Developer edition and higher). It covers four areas:
Current architecture: a visual model of how your project is structured and how its components depend on one another.
Intended architecture: define which components may and may not depend on each other, and SonarQube raises deviations as standard issues.
Deviations, flaws, and smells: structural problems such as tangles, oversized components, and split responsibilities are detected automatically.
Remediation workflow: architecture issues follow the same triage and fix workflow as code quality issues, integrated with quality profiles and quality gates.
Supported languages are Java, C#, JavaScript, TypeScript, and Python. For more information, see Architecture.
IPv6 support for DCE Docker and Helm deployments
SonarQube Server Data Center Edition now supports IPv6-only and dual-stack network environments for both Docker and Kubernetes (Helm) deployments. The Docker entrypoint automatically discovers the node address across IPv4 and IPv6, and a new SONAR_CLUSTER_NODE_HOST environment variable lets you set a static IPv6 address on dual-stack networks where you want to run the DCE cluster exclusively over IPv6. The Helm chart works on IPv6-only and IPv6-primary dual-stack clusters without any chart-level configuration, using the pod's primary IP family as exposed by the Kubernetes downward API. For more information, see Customizing the DCE Helm chart.
EU Cyber Resilience Act security report
Available in Enterprise edition and higher.
A new EU Cyber Resilience Act (CRA) security report maps detected findings to the CRA requirements, and a corresponding entry is available in the Security Category facet. This helps teams demonstrate compliance with the EU CRA from within SonarQube Server. For more information, see Cyber Resilience Act.
Advanced Security
Available as part of SonarQube Advanced Security license for Enterprise edition and higher.
Bulk actions for dependency risks
You can now select multiple dependency risks and update them as a group, applying a status change to many risks in a single action instead of one at a time. This speeds up triage when the same decision applies across several dependency risks. For more information, see Reviewing and fixing dependency risks.
Languages
Note: Security hotspots rules are being changed to vulnerabilities (in Standard Experience) or security issues (in MQR Mode). Existing security hotspots remain visible on the Security Hotspots page.
Gosu (new)
SonarQube Server now supports Gosu, the language used with the Guidewire platform in the insurance sector. Support includes parsing, syntax highlighting, and metrics, along with an initial set of code quality rules.
New rules:
S100: Function names should comply with a naming convention
S101: Class names should comply with a naming convention
S103: Lines should not be too long
S104: Files should not have too many lines of code
S105: Tabulation characters should not be used
S107: Functions should not have too many parameters
S108: Nested blocks of code should not be left empty
S117: Local variable and function parameter names should comply with a naming convention
S121: Control structures should use curly braces
S122: Statements should be on separate lines
S126: "if ... else if" constructs should end with "else" clauses
S131: "switch" statements should have "default" clauses
S134: Control flow statements "if", "for", "while", "switch" and "try" should not be nested too deeply
S138: Functions should not have too many lines of code
S1066: Mergeable "if" statements should be combined
S1125: Boolean literals should not be redundant
S1134: Track uses of "FIXME" tags
S1135: Track uses of "TODO" tags
S1143: Jump statements should not occur in "finally" blocks
S1144: Unused "private" methods should be removed
S1145: Useless "if(true) {...}" and "if(false){...}" blocks should be removed
S1151: "switch case" clauses should not have too many lines of code
S1172: Unused function parameters should be removed
S1186: Functions should not be empty
S1192: String literals should not be duplicated
S1206: "equals" and "hashCode" should be overridden in pairs
S1313: IP addresses should not be hardcoded
S1451: Track lack of copyright and license headers
S1479: "switch" statements should not have too many "case" clauses
S1656: Variables should not be self-assigned
S1764: Identical expressions should not be used on both sides of a binary operator
S1821: "switch" statements should not be nested
S2260: Gosu parser failure
S3776: Cognitive Complexity of functions should not be too high
S8708: URL validation should use established libraries instead of string operations
S8711: String literals should be used with DisplayKey.get()
S8712: Multiple "where()" calls should be combined into a single filter
S8713: Deprecated "intersect()" methods should not be used
S8724: BigDecimal and BigInteger should use literal suffixes
S8725: Private variables should be prefixed with an underscore
S8726: Public variables should start with an uppercase letter
S8727: "new Date()" should not be used in Guidewire applications
S8729: Use "hasMatch()" instead of "where().count()" when checking for element existence
S8730: "@PostOnChange" annotations should specify a target property
S8731: Query "select()" should specify columns instead of returning all fields
S8734: Transaction bundles should not be committed explicitly inside "runWithNewBundle" blocks
S8735: XML content should be logged instead of printed to standard output
S8736: Trailing comments should be aligned within code blocks
S8737: Comment syntax should match comment length
S8738: Messaging plugin lifecycle event handlers should not be empty
S8739: Properties should be used instead of getter/setter methods
S8740: Threads should not be created manually in container-managed environments
For more information, see Gosu.
C# and VB.NET
The .NET analyzer is updated across the cycle with a new rule and rule improvements.
New rule:
S8717: Multiple "[Key]" attributes should not be used to define a composite key
Examples of modified rules:
S108: Nested blocks of code should not be left empty
S1133: Deprecated code should be removed
S1135: Track uses of "TODO" tags
S2629: Logging templates should be constant
S2696: Instance members should not write to "static" fields
S3251: Implementations should be provided for "partial" methods
Java
The Java analyzer adds new rules across the cycle, including rules for the Java date and time APIs and rules covering Quarkus, JPA, and Bean Validation patterns. Analysis no longer fails when sonar.java.binaries is missing; a warning is logged instead.
New rules:
S8220: Conversions between local and timezone-aware types should use explicit timezone handling
S8688: Time-based .now() methods should specify a ZoneId or a Clock
S8692: The system clock should not be used in unit tests
S8694: DayOfWeek and Month Enums should be used instead of numeric values
S8695: Redundant time instantiation patterns should be simplified
S8696: Value-based types should be compared using their value
S8700: Do not perform date and time arithmetic on DST unaware types
S8908: Methods annotated with "@CacheResult" should not return void
S8909: "CacheKeyGenerator" implementations should be instantiable by the framework
S8910: Interfaces annotated with "@Mapper" should contain at least one "@DaoFactory" method
S8911: Methods annotated with "@Startup" should be non-static, non-producer, and parameter-free
S8912: Custom CredentialsProvider implementations should be annotated with "@Unremovable"
S8913: REST Data with Panache resource interfaces should not have implementation classes
S8947: JPA entity classes should not be final
S8948: "@OneToMany" relationships should use "mappedBy" or "@JoinColumn"
S8954: Bean Validation constraints should not be placed on static fields
For more information, see Java.
JavaScript/TypeScript/CSS
The JavaScript and TypeScript analyzer adds sonar-resolve support and detection of generated code and ships new CSS rules. New rules also target testing practices, utility-library usage such as Lodash, and Vue.
New rules:
S2925: No fixed waits in tests
S5976: Similar tests should be grouped in a single Parameterized test
S8782: Lifecycle hooks should be declared before test cases
S8783: Forced browser interactions should not bypass actionability checks
S8784: Assertions should be placed inside test cases or hooks
S8785: Test suite callbacks should be synchronous functions
S8907: Replace Lodash/underscore.js methods with native APIs
S8927: Default imports from modular utility libraries should not be used
S8950: Vue props with a default value should not be required
S8951: Vue component props should not be mutated directly
S8959: UI test debug commands should not be used
S8961: Vue component events should be explicitly declared
S8962: Vue ref and shallowRef calls should be typed
S8967: Inline snapshots should not contain interpolations
For more information, see JavaScript/TypeScript/CSS.
Python
The Python analyzer adds new rules for testing, the Pydantic library, and web scraping with Beautiful Soup, and reduces false positives. Issues classified as main-code issues are no longer raised on test files.
New rules:
S2187: Test files should contain at least one test case
S3415: Assertion arguments should be passed in the correct order
S5778: Only one method invocation is expected when testing runtime exceptions
S5779: Assertions should not be made within the try block of a try-except catching AssertionErrors
S5863: Assertions should not be given twice the same argument
S5958: Tests should check which exception is thrown
S5976: Similar tests should be grouped in a single Parameterized test
S8714: Dedicated exception assertions should be used instead of "try-catch" with "fail()"
S8900: Deprecated Beautiful Soup methods should be replaced with their modern equivalents
S8903: HTML elements should be created with "new_tag()" instead of inserting raw HTML strings
S8904: BeautifulSoup elements should be checked for None before accessing
S8905: BeautifulSoup parser should be explicitly specified
S8906: CSS selectors should be used when selecting elements with multiple required classes
S8963: Pydantic models should not use multiple inheritance with conflicting configurations
S8966: pydantic-core serialization calls should provide fallback handlers
S8971: "SkipValidation" should not be combined with validation constraints
S8973: Double leading underscores should not be used for private attributes in Pydantic models
S8974: "json_schema_input_type" should not be used with "mode='after'" field validators
For more information, see Python.
HTML
A new rule lets you enforce the HTML lang attribute against a configured list of allowed language codes.
New rule:
S8687: Enforce the HTML "lang" attribute against a configured list of ISO 639-1 language codes
For more information, see HTML.
PL/SQL and T-SQL
PL/SQL and T-SQL properties are now grouped under a shared SQL category in the configuration UI. For more information, see PL/SQL and T-SQL.
PostgreSQL (new)
SonarQube Server now analyzes the PostgreSQL SQL dialect, including PL/pgSQL, and ships an initial set of rules covering correctness, security, and maintainability of PostgreSQL code.
New rules:
S8788: Set-returning functions should terminate with a bare "RETURN" after "RETURN NEXT" or "RETURN QUERY"
S8933: Composite primary keys should not contain more than four columns
S8934: Schemas should not grant the CREATE privilege to PUBLIC
S8935: Format strings should have properly terminated conversion specifiers
S8937: Format placeholders in "format()" should use 1-based indexing
S8938: Indirect-width format specifiers should include a dollar sign after the width position
S8939: PL/pgSQL protected automatic variables should not be modified
S8942: Identifiers should not use reserved SQL keywords
S8943: Unused arguments in PostgreSQL "format()" calls should be removed
S8944: PostgreSQL "format()" calls should provide enough arguments for all placeholders
S8945: Transaction control statements should not be used inside exception handling blocks
For more information, see PostgreSQL.
Secrets
The following improvements have been made:
The list of file extensions excluded from secret detection, previously hard-coded, is now customizable.
Test-file detection, performance, and logging have been improved.
The known-fake-secret filter discards secret candidates that match a rule's list of fake or placeholder values (repeated characters, placeholder tokens, documentation examples). The new sonar.secrets.disableKnownFakeSecretFilter property (default false) turns this filter off when set to true, reporting those candidates as low-confidence matches, which is useful when evaluating secret detection on a benchmark project. For more information, see Secrets.
Infrastructure as Code
The Infrastructure as Code analyzer improves file-predicate handling with a new cache and adds new rules, including new Logic Apps rules for Azure Resource Manager.
New rules:
S8793: RDS database instances should not be publicly accessible
S8847: Azure Key Vaults should enable purge protection
S8857: GCP Cloud Storage buckets should enforce uniform bucket-level access
2026.3
AI capabilities
Embedded Model Context Protocol (MCP) for SonarQube Server
SonarQube Server can now host a SonarQube MCP server as an extension, exposing a /mcp reverse-proxy endpoint that lets AI coding agents (Claude, Cursor, Copilot, and others) query your SonarQube Server instance directly. The hosted MCP server is installed on SonarQube Server, simplifying management by requiring only a single MCP server URL to access.
Administrators can toggle the server, configure the upstream URL, and tune health-check behavior through sonar.mcp.* properties in your sonar.properties file. Health-check status is surfaced through the standard SonarQube Server monitoring endpoints. See the MCP server https://docs.sonarsource.com/sonarqube-developer-tools/sonarqube-mcp-server/setup/sonarqube-server-hosted page for more information.
Sonar agentic AI quality profile
A new built-in quality profile, Sonar agentic AI, is now available for Java, JavaScript/TypeScript, and Python. The profile selects rules tuned for code produced by AI coding agents, focusing on the failure modes and recurring patterns most often introduced by agentic workflows.
Monitoring alerts for SonarQube Server administrators
Monitoring alerts help system administrators detect performance problems in a SonarQube Server instance. They surface signs of degraded behavior early, so administrators can investigate before users start reporting slow or failed analysis results. Their purpose is to show that the instance is experiencing a performance problem and that administrator attention is needed. See Monitoring alerts for more information.
GitLab authentication and provisioning improvements
This major overhaul removes critical friction points for large-scale enterprise GitLab customers with tens of thousands of users and projects. It officially maps the new GitLab Planner and Security Manager roles to SonarQube's default Reporter permissions. It also vastly optimizes GitLab JIT provisioning times and provides a new Allow all groups option.
Configurable server-side cache expiration
The expiration duration for the server-side cache used for faster PR analysis is now configurable, allowing instances to tune cache lifetimes to match their analysis cadence and infrastructure. Set the number of days for the sonar.dbcleaner.daysBeforeDeletingScannerCache property in the SonarQube Server UI at Administration > General Settings > Housekeeping.
Improved license renewal experience
Licenses issued for a specific edition remain valid for all editions below that licensed tier, so you have time to upgrade when renewing or changing your SonarQube Server edition. See License administration for more information.
Advanced Security
Available as part of SonarQube Advanced Security license for Enterprise edition and higher.
Vulnerability Exploitability Exchange (VEX) export
Generate and download vulnerability reports in the CycloneDX 1.6 VEX format. This extends existing risk reporting by automatically compiling a list of vulnerabilities affecting a project or portfolio, translating their current status from dependency risk metrics, and pulling in the specific engineering justifications based on status-change comments. See Reviewing and fixing dependency risks for more information.
Change in scope of Software Bills of Materials
Starting with SonarQube Server 2026.3, Software Bills Of Materials (SBOMs) generated by SonarQube Advanced Security only include production dependencies or dependencies that end up in your released application by default. If you need to include all dependencies, including development dependencies, you can generate a SBOM via the API by passing onlyProductionScope=false.
Languages
C#
Four new rules help teams adopt C# 14 cleanly by flagging identifier conflicts with the new contextual keywords. See C# for more information.
New rules:
S8367: Identifiers should not conflict with the "field" keyword in C# 14
S8368: "extension" identifiers should be escaped to avoid contextual keyword conflicts
S8380: Return types named "partial" should be escaped with "@"
S8381: "scoped" should be escaped when used as a type name in lambda parameters
Cobertura coverage format for C#
The C# analyzer now accepts Cobertura-formatted coverage reports, passed via the sonar.cs.cobertura.reportsPaths parameter. This complements the existing dotCover, OpenCover, and Visual Studio coverage formats and removes the need to convert Cobertura output before importing it. See .NET test coverage for more information.
Go
Go test files are now analyzed only by checks applicable to test files, removing irrelevant findings on _test.go files. See Go for more information.
Groovy
New Groovy rules ship in this release: 17 for the core Groovy language and 23 specific to Jenkins pipelines. See Groovy for more information.
New rules for the Groovy language:
S107: Functions should not have too many parameters
S122: Statements should be on separate lines
S126: "if ... else if" constructs should end with "else" clauses
S134: Control flow statements "if", "for", "while", "switch" and "try" should not be nested too deeply
S138: Functions should not have too many lines of code
S1067: Expressions should not be too complex
S1125: Boolean literals should not be redundant
S1134: Track uses of "FIXME" tags
S1135: Track uses of "TODO" tags
S1145: Useless "if(true) {...}" and "if(false){...}" blocks should be removed
S1151: "switch case" clauses should not have too many lines of code
S1192: String literals should not be duplicated
S1479: "switch" statements should not have too many "case" clauses
S1821: "switch" statements should not be nested
S1862: Related "if-else if" statements should not have the same condition
S3923: All branches in a conditional structure should not have exactly the same implementation
S4663: Multi-line comments should not be empty
New rules for Jenkins pipelines:
S8327: Jenkins pipeline scripts should use pipeline steps instead of direct file I/O operations
S8351: Input statements should be wrapped with timeouts and placed outside agent blocks
S8353: Pipeline parameters should not use environment variables in default values
S8355: Variables containing complex objects should not be declared in environment blocks
S8356: Pipeline parameter definitions should not reference locally-defined environment variables
S8357: Methods should use @NonCPS annotation to avoid CPS transformation issues
S8358: Pipeline steps should not be called from "@NonCPS" methods
S8359: Methods with closure parameters should not have ambiguous overloads
S8360: "getItemByFullName" should be used to access Jenkins jobs in folders
S8361: String split results should not be accessed by index without bounds checking
S8364: JUnit step should specify test results file pattern
S8365: Temporary files should be deleted after use in Jenkins pipelines
S8366: Script-level variables should use "@Field" annotation instead of binding variables
S8524: Try-catch blocks should be wrapped in "script" blocks in declarative pipelines
S8525: Scripted code should be wrapped in "script" blocks within declarative pipeline stages
S8526: SCM checkouts should use dedicated steps instead of shell commands with credentials
S8527: "credentials()" should be used instead of "withCredentials" in environment sections
S8531: Declarative and Scripted Pipeline syntax should not be mixed
S8535: Groovy script strings should not be duplicated
S8536: Jenkins parallel steps should use named arguments
S8538: PATH modifications in "environment" blocks should use "$PATH" instead of "${env.PATH}"
S8539: Choice parameters should be passed as strings when calling Jenkins jobs
S8540: GitHub source blocks in Jenkins multibranch pipelines should include an explicit "id" field
Java
The Java analyzer ships a new rule alongside bug fixes and stability improvements. See Java for more information.
New rule:
S3706: "stream" should not be used for Collection "forEach" calls
JavaScript/TypeScript
One rule has been deprecated. See JavaScript/TypeScript/CSS for more information.
Deprecated rule:
S5042: Expanding archive files should not be done without controlling resource consumption
Package manager analyzer
New rules detect missing lock files across multiple languages:
S8564–S8571: (JavaScript, Python, Go, PHP, Ruby, Java and Kotlin, Rust, Dart) dependency lock file should be committed to source control
New rule that detects Unicode Variation Selectors:
S8522: Consecutive Unicode Variation Selectors should not be used
PHP
One rule has been deprecated. See PHP for more information.
Deprecated rule:
S4828: OS processes should not be signaled without validation
PowerShell
SonarQube Server now includes coverage of PowerShell scripts. See PowerShell for more information.
New rules:
S3776: Cognitive Complexity of functions should not be too high
S8429: Cmdlets should be invoked with all mandatory parameters
S8620: Lines should not end with trailing whitespace
S8621: Pipeline statements spanning multiple lines should use consistent indentation
S8622: "!" should not be used for logical negation
S8624: "HelpMessage" parameter attributes should not be null or empty
S8625: Functions should not shadow built-in PowerShell cmdlets
S8626: Automatic variables should not be assigned to
S8628: Hash algorithms MD5 and SHA-1 should not be used
S8631: Parameter sets should have at most one parameter accepting pipeline input by value
S8633: DSC resource functions should have identical parameters
S8634: DSC resources should implement all required functions
S8637: Reserved common parameters should not be redefined in advanced functions
S8638: Deprecated WMI cmdlets should not be used
S8640: Switch parameters should not default to "$true"
S8641: "$null" should be placed on the left side of comparison operators
S8642: Cmdlets, parameters, keywords, and operators should use consistent casing
S8647: Credentials should not be sent over unencrypted connections
S8649: Cmdlet aliases should not be used in scripts
S8652: Credential parameters should use the PSCredential type
S8653: DSC class "Test" methods should return boolean values
S8657: Catch blocks should not be empty
S8659: "Invoke-Expression" should not be used
S8661: Parameters should have only one type specifier
S8664: Mandatory parameters should not have default values
S8666: Lines should not end with a backtick followed by whitespace
S8667: Module manifests should use "RootModule" instead of deprecated "ModuleToProcess"
S8669: DSC class "Set" methods should return void
S8672: Functions accepting pipeline input should use a "process" block
S8673: Computer names should not be hardcoded
S8675: Function and cmdlet names should not use reserved words or reserved characters
S8677: Functions should not use "Write-Host" unless they use the "Show" verb
Python
Python collections rules
New rules for Python collections target readability, correctness, and performance issues, guiding teams toward more idiomatic constructs such as min() and max(), direct membership tests on dictionaries, next(iter(...)), itertools.chain.from_iterable(...), distinct loop variables, and simpler set operations. See Python for more information.
New rules:
S8492: "set.discard()" should be used instead of checking membership before removal
S8493: "StopIteration" should not be raised inside generators
S8503: Membership tests should not use empty collections
S8510: Loop variables should not be reused in nested loops
S8512: Class fields should not be defined multiple times
S8517: "sorted()" should not be used with indexing to find minimum or maximum values
S8519: "list(...)[0]" should not be used to get the first element
S8520: "sum()" should not be used with an empty list to concatenate lists
S8521: Dictionary membership tests should not explicitly call ".keys()"
Python object-oriented programming rules
Seven new rules target common object-oriented pitfalls, helping teams catch broken inheritance hierarchies, unsafe dataclass defaults, incomplete comparison logic, missing property returns, invalid __slots__ assignments, duplicate base classes, and inconsistent tuple-return contracts. See Python for more information.
New rules:
S8494: Attributes should only be assigned if they are declared in "slots"
S8495: Functions should return tuples of consistent length
S8500: Comparison methods should be defined completely
S8504: Property methods should have a return statement
S8509: Classes should not inherit from the same base class multiple times
S8511: Multiple inheritance should not create Method Resolution Order (MRO) conflicts
S8514: Dataclass attributes should use type annotations and "default_factory" for mutable defaults
Python data structures and operations rules
Five new rules focused on data structures and operations help teams catch subtle but high-impact bugs around enums, dataclasses, dispatch decorators, shared mutable defaults, and iterator reuse. See Python for more information.
New rules:
S8505: @singledispatch and @singledispatchmethod should not be confused
S8508: Mutable default values should not be used with dict.fromkeys() or ContextVar()
S8516: Group iterators from itertools.groupby should not be reused
S8685: Function calls should not be used as default values in dataclass attributes
RPG
Four new RPG rules ship in this release. The analyzer now supports multiline issues, allowing single findings to span more than one line of code. See RPG for more information.
New rules:
S1896: "INZ()" should not be used on module-level standalone fields in TSR programs
S2033: Library names should not be hard-coded
S2284: Calculations should use free-form syntax
S2794: Result data structures should be used for file I/O
Ruby
Ruby analysis has been improved, delivering enhanced analysis performance. See Ruby for more information.
Rust
The Rust analyzer now supports the following Clippy rules when importing a Clippy report. See Rust for more information.
Supported Clippy rules:
disallowed_fields
duration_suboptimal_units
manual_checked_ops
manual_take
unnecessary_trailing_comma
Infrastructure as Code
The IaC analyzer adds multi-document Helm chart support and improves Bicep parsing. A new set of supply-chain-focused rules for Shell, Docker, Azure Pipelines, and GitHub Actions covers risky package-manager and CI patterns. See Supported languages for more information.
Example new rules:
S6505: Allowing shell scripts execution during package installation is security-sensitive
S7694: Swift dependencies should be locked to verified versions
S8482: Avoid executing downloaded artifacts without verification
S8531: Declarative and Scripted Pipeline syntax should not be mixed
S8543–S8550: (JavaScript, Python, Go, PHP, Ruby, Java and Kotlin, Rust, Dart) dependencies should be locked to verified versions
2026.2
New layout and navigation for SonarQube Server
The SonarQube Server UI has a refreshed layout and navigation. The horizontal top menu has been replaced with an intuitive vertical sidebar, introducing a new context switcher that allows users to instantly jump between enterprises, organizations, portfolios, and projects without losing their place.
In-code issue resolution (sonar-resolve)
New in-code annotations let you resolve specific issues by rule, with a mandatory comment and status (accepted or false positive), so deviations stay visible in SonarQube’s UI and auditable rather than being blindly suppressed. This structured alternative to NOSONAR helps teams comply with standards such as MISRA C++:2023 and reduces the risk of accidentally hiding critical issues on the same line. See Editing issues for more information.
Key Capabilities:
Set resolution status (accept or fp for false-positive) directly in the code.
Administrative control via global and project-level settings.
Supported languages: C, C++, Objective-C
AI CodeFix enhancements
Improvements to AI CodeFix configuration to make it model-agnostic, enabling better flexibility in AI-powered code fix suggestions. This feature is available in Enterprise edition and higher.
License management improvements
SonarQube Server now automatically refreshes SonarQube license every 12 hours for instances using online activation, ensuring immediate access to new features and LOC limit updates without manual intervention. See Online license management for details.
Advanced Security
Available as part of SonarQube Advanced Security license for Enterprise edition and higher. See Advanced Security for more information. See Introduction to Advanced Security for more information.
Dependency risks in security reports
Sonar security reports now include a Dependency risk column. This weaves Software Composition Analysis (SCA) data directly into application and portfolio-level reports in both the SonarQube Server UI and exported PDFs. See Security reports for details.
Risk report and SBOM in regulatory reports
Project regulatory reports now include both a risk report and a software bill of materials SBOM that you can download from your projects. See Regulatory reports for details.
ASAST Configurations for the Python Top 1K
We are expanding Advanced Static Application Security Testing (ASAST) support with the top 1,000 most utilized libraries in the Python ecosystem.
Languages
Apex code quality rules
SonarQube Server 2026.2 expands Apex support with 23 new code quality rules, providing enhanced coverage for Salesforce developers. Apex support is available in Enterprise edition and higher. See Apex for more information.
S1213: The members of an interface or class declaration should appear in a pre-defined order
S1659: Multiple variables should not be declared on the same line
S7951: Database SaveResult objects should be checked for errors
S7965: Future methods should not accept sObjects or custom objects as parameters
S7972: Apex cursor fetch should use small chunk sizes to avoid governor limits
S7994: AuraEnabled methods should be static when they don't require instance state
S7999: Email operations should include proper error handling
S8000: Test classes should create required test data within the test
S8001: SOQL LIKE clauses should not use leading wildcards
S8008: Encryption keys should not be hardcoded
S8020: Server actions that retrieve data should be marked as cacheable
S8028: Future methods should not be called from batch or queueable contexts
S8032: Database.Stateful should only be used when state retention is needed
S8035: Change Data Capture event objects should follow the correct naming convention
S8041: Apex callouts should implement retry logic for reliability
S8044: FormulaEval.FormulaBuilder should be properly configured with null checks, type safety, and return type
S8125: Field-level permissions should be checked before accessing fields
S8130: Retired Salesforce API versions should not be used
S8451: Schema describe operations should not be called inside loops
S8452: Classes should override both equals and hashCode or neither
S8453: Test assertions should include descriptive messages
S8455: SObject describe calls should use deferred loading
S8456: Annotations should use PascalCase naming convention
Improved Go analyzer performance
Go analyzer is now 30 times faster. New rules include the same rules as listed for Apex above.
Java 25 support
SonarQube 2026.2 introduces error-free parsing and deep semantic analysis for Java 25 LTS, the first long-term support release since JDK 21. We've added critical rules targeting new features like Scoped Values (JEP 506), Flexible Constructor Bodies (JEP 513), and Module Imports (JEP 511). Crucially, these rules are designed to catch syntactically valid but semantically broken code generated by AI assistants trained on outdated preview APIs. See Java for more information.
Examples of new rules:
S1128: Redundant imports should be removed
S3051: Main methods should be used only as program entry point
S8432: "ScopedValue.where" results should not be ignored
S8433: Validation logic should be placed in constructor prologue when possible
S8444: Validation and data preparation logic before super() should not bloat constructor
S8445: Group import declarations by specificity
S8446: Only one "main" method should be present
S8447: Initialize subclass fields before super() when superclass constructor may call overridable methods
S8450: Use IO.readln() for console input instead of BufferedReader boilerplate
S8465: "ScopedValue" instances should be assigned to a stable reference
S8469: Use IO.readln(String prompt) instead of IO.print followed by IO.readln()
JavaScript / TypeScript security rules
Six new security rules have been added for JavaScript / TypeScript. See JavaScript/TypeScript/CSS
S5335: Imports should not be vulnerable to injection attacks
S6547: Environment variables should not be defined from untrusted input
S5496: Server-side templates should not be vulnerable to injection attacks
S6641: Connection strings should not be vulnerable to injection attacks
S6549: Accessing files should not lead to filesystem oracle attacks
S7518: Privileged prompts should not be vulnerable to injection attacks
Kotlin 2.3.10 support
Upgraded the Kotlin analyzer for version 2.3.10 support.
Python
Python Django framework rules
New rules specifically targeting Django best practices and common pitfalls for web developers. See Python for more information.
Rules added:
S8437: Class-Based Views should override "get_context_data" correctly
S8438: Django view functions should declare URL parameters explicitly
S8439: Django view functions should include all URL parameters in their signature
S8440: Querysets should use "select_related()" or "prefetch_related()" to avoid N+1 queries
S8443: Django Command classes should inherit from BaseCommand
S8486: Django middleware should call super().init() with appropriate parameters
Python Flask rules
Flask services get dedicated rules to harden configuration, routing, and error handling, focusing on security and correctness of HTTP behavior. See Python for more information.
Related rules:
S6863: Flask error handlers should set HTTP status code
S6965: Flask REST API actions should be annotated with an HTTP verb attribute
S8370: Query parameters should not be used to carry body data in POST requests
S8371: HTTP headers should be accessed safely to avoid KeyError
S8374: Flask class-based view decorators should be applied using the decorators attribute
S8375: Flask preprocess_request() return values should be handled
S8385: send_file should specify mimetype or download_filename
S8388: Flask applications should not bind to all network interfaces
Python FastAPI rules
FastAPI projects now get framework‑aware rules around routing, Pydantic models, dependencies, and documentation, aimed at catching typical FastAPI mistakes early. See Python for more information.
Related rules:
S8389: File upload endpoints should use Form() with Pydantic
S8392: FastAPI applications should not bind to all network interfaces
S8396: Optional Pydantic fields should have explicit default values
S8397: FastAPI applications should be passed as import strings when using reload
S8400: Endpoints returning 204 should have an empty body
S8401: Child routers should be included before parent router registration
S8405: TestClient requests should use the content parameter
S8409: Endpoints should not specify redundant response_model parameters
S8410: Dependencies should use Annotated type hints
S8411: Path parameters should be included in route function signatures
S8412: Generic route decorators should not be used
S8413: Router prefixes should be defined during APIRouter initialization
S8414: CORSMiddleware should be added last in the middleware chain
S8415: HTTPException responses should be documented in endpoint metadata
Groovy language support (beta)
Initial support for Groovy language with 30+ code quality rules, enabling analysis of Groovy-based build files and scripts. See Groovy for more information.
Related rules:
S8289: File operations should specify charset encoding
S8298: "@TimedInterrupt" should not be used on static methods
S8299: AST transformation classes should be annotated with "@CompileStatic"
S8322: Simple "@Grab" annotations should use shorthand notation
S8268: Thread.sleep() should not be used in loops for busy waiting
S8287: Test methods should contain assertions
S8323: Property names should use camelCase
S8314: Static imports should appear before regular imports
S8309: Use appropriate sorting methods to avoid mutation confusion
S8285: Method names should follow camelCase naming conventions
S8269: "wait()" calls should be inside "while" loops
S8315: Empty strings should not be used for type conversion
S8308: Elvis operator should be used for null-safe operations and ternary simplification
S8304: Duplicate import statements should be removed
S8275: Null checks should use correct logical operators
S8320: GString expressions should not be used as map keys
S8326: Range methods should be used appropriately
S8311: Method names should not use reserved keywords
S8272: Classes with a "clone()" method should implement "Cloneable"
S8303: Star imports should be replaced with explicit imports
S8307: Semicolons should be omitted in Groovy
S8306: Control structures should use braces
Ruby (beta)
There are eight ruby rules in beta and two that have been removed. See Ruby for more information.
New beta rules:
S8418: Unused method and block parameters should be removed or prefixed with underscore
S8419: Function parameters should not be immediately reassigned
S8421: Underscore-prefixed variables should not be used
S8422: Trailing underscores in multiple assignment should be removed
S8423: Parameter default values should not reference themselves
S8424: Constants should not be reassigned
S8425: Constants should be explicitly scoped to avoid ambiguous resolution
S8426: Variables should not be assigned only to be implicitly returned
Removed rules
S1854: Unused assignments should be removed
S7819: Variables and methods should be accessible in their usage context
2026.1
AI and mobile compliance reporting
Extends our regulatory coverage to include critical AI and Mobile security standards such as OWASP Top 10 for LLM and OWASP MASVS for project security reports. This feature is available in the Enterprise edition and above. See Security-related rules for more information.
Feedback mechanism for self-hosted LLMs
Improves the success rate of generating valid AI CodeFix suggestions from self‑hosted LLMs.
JFrog Evidence Collection with SonarQube Server
This integration provides a single, verifiable audit trail if you use both SonarQube and JFrog with strict audit trail and compliance requirements. SonarQube analysis results are automatically signed and directly attached to your JFrog packages to create a single, verifiable source of truth. You no longer have to jump between tools to prove your code meets security standards. Everything you need for a rigorous audit is now visible within the JFrog Evidence Collection interface. This feature is available in the Enterprise edition and above. See JFrog Evidence Collection for more information.
SonarQube Advanced Security
This feature is available in the Enterprise edition and above.
Malicious package detection
Receive blocker-level alerts if a dependency matches publicly known datasets of known malicious packages. See Advanced Security for more information.
Quality gate fudge factor improved
To avoid overly strict enforcement of small changes, the quality gate ignores coverage and duplication conditions for very small sets of new code. See Changing instance's default quality gate for more information.
Languages
Cobol
Adds support for parsing additional language constructs and includes fixes for crashes and false positives for COBOL. Related rules include:
S3938: Track uses of forbidden statements
S1725: Open files should be closed explicitly
S1574: Data items should be initialized with data of the correct type
S1289: Unused data item blocks should be removed
IaC
The analysis of Infrastructure as Code (Ansible, Azure Resource Manager, CloudFormation, Docker, Kubernetes, Terraform, Shell, GitHub Actions) has been improved.
Helm templates are now evaluated even if values.yaml is missing.
The following rules have been added:
S6437: Credentials should not be hard-coded
S7638: ACTIONS_ALLOW_UNSECURE_COMMANDS should not be used
S8232: Workflows should not rely on unverified GitHub context values to trust events
S8233: Write permissions should be defined at the job level
S8262: Artifacts should not contain secrets
S8263: GitHub Action invocations should not be vulnerable to parameter injection attacks
S8264: Read permissions should be defined at the job level
JCL
A new leaveFile API is available for custom rules for JCL language, giving rule authors more control over how files are processed and reported.
.NET 10 and C# 14 support
Empowers .NET teams to adopt the Long Term Support (LTS) release of .NET 10 and C# 14 immediately, ensuring their analysis remains accurate, performant, and free of false positives associated with new language constructs. See VB.NET and C# for more information.
Related rules:
S1121: Assignments should not be made from within sub-expressions
S1144: Unused private types or members should be removed
S2225: "ToString()" method should not return null
S2292: Trivial properties should be auto-implemented
S2325: Methods and properties that don't access instance data should be static
S2583: Conditionally executed code should be reachable
S2589: Boolean expressions should not be gratuitous
S2692: "IndexOf" checks should not be for positive numbers
S2953: Methods named "Dispose" should implement "IDisposable.Dispose"
S2970: Assertions should be complete
S3063: "StringBuilder" data should be used
S3264: Events should be invoked
S3398: "private" methods called only by inner classes should be moved to those classes
S3459: Unassigned members should be removed
S3877: Exceptions should not be thrown from unexpected methods
S3928: Parameter names used into ArgumentException constructors should match an existing one
S4545: "DebuggerDisplayAttribute" strings should reference existing members
S7039: Content Security Policies should be restrictive
PHP
Reduces false positives on several rules and cleans up build and dependency infrastructure for PHP. Related rules:
S1155: "empty()" should be used to test for emptiness
S1172: Unused function parameters should be removed
S2699: Tests should include assertions
S1068: Unused "private" fields should be removed
Scala
Include fixes to false positives and negatives for Scala on the following rules:
S1192: String literals should not be duplicated
S126: "if ... else if" constructs should end with "else" clauses
Secrets
Secrets rules have been improved to reduce the detection of false positives and the following rule have been added:
S6418: Hard-coded secrets are security-sensitive
S2068: Hard-coded passwords are security-sensitive
S7552: SMTP credentials should not be disclosed
S8350: xAI API keys should not be disclosed
VB6
Fixes parse errors and line count for VB6. Related rules:
S138: Subs and functions should not have too many lines
S1151: "Case" clauses should not have too many lines
Update notes
This section contains notes about breaking changes and important updates to be aware of before updating. If you’re updating from the previous LTA, see 2026.1 LTA to LTA release notes
2026.4
SonarQube Server 2026.4 includes Elasticsearch 9.x
SonarQube Server 2026.4 and later includes Elasticsearch 9 (9.4.3), which requires read and write access to the /tmp directory. This is a requirement from Elasticsearch itself and cannot be disabled. For more information and a solution, see Elasticsearch filesystem requirements.
Upgrading to Elasticsearch 9 triggers a full re-index. Because Lucene 10 cannot read the previous on-disk format, the index is rebuilt from scratch into a fresh es9 subdirectory. Your configured index location is preserved, but its contents are regenerated.
2026.3
Update to Microsoft SQL JDBC Auth 13.4.0
To use integrated security in a Microsoft SQL database, upgrade to the Microsoft SQL JDBC Auth 13.4.0 package. See Installing database for more information.
2026.1
Java requirements for SonarQube Server runtime
The SonarQube Server runtime now requires Java Development Kit (JDK). The previous requirement of a Java Runtime Environment (JRE) is no longer sufficient, and a full JDK is required.
Added Support for Java 25 in addition to Java 21.
Removed support for Java 17.
See Software requirements for more details.
PostgreSQL support
Support for PostgreSQL versions 14 through 18 is now available, enabling deployments using the most recent PostgreSQL release. PostgreSQL version 13 is not supported anymore. See Installing database for more information.
Kubernetes and Openshift support
Supported Kubernetes Versions: From 1.32 to 1.35. Support for versions 1.30 and 1.31 has been removed.
Supported Openshift Versions: From 4.17 to 4.20. Support for versions 4.11 to 4.16 has been removed.
Support for MSSQL server
Supported MSSQL server is now 2022 (MSSQL Server 16.0); 2019 (MSSQL Server 15.0); 2017 (MSSQL Server 14.0). Support for 2016 MSSQL Server 13.0 support has been removed. See Installing database for more information.
SonarQube Server includes Elasticsearch 8.x
SonarQube Server 2026.1 LTA and later includes Elasticsearch 8.x, which requires read and write access to the /tmp directory. This is a requirement from Elasticsearch itself and cannot be disabled. For more information and a solution, see Elasticsearch filesystem requirements.
Deprecations and removals
This section contains information on the deprecation and removal of SonarQube Server features and API endpoints. See the Deprecation policy for more information.
2026.4
Deprecation of Security Hotspots
To simplify the classification of findings, security hotspots are deprecated. Rules that previously raised security hotspots now raise vulnerabilities (in Standard Experience) or security issues (in MQR Mode) instead. Existing security hotspots remain visible on the Security Hotspots page.
Java 17 support for SonarScanners has been removed
Support for Java 17 in SonarScanners has been removed in SonarQube Server 2026.4. Starting with version 2026.4, Java 21 is required. If you use JRE auto-provisioning (enabled by default on supported scanners), no action is required as Java requirements are managed automatically. However, if auto-provisioning is disabled or unsupported, you must upgrade to Java 21 or newer. For more information, see General requirements.
2026.1
Java 17 not supported any more
Java version 21 is the minimum version required to run SonarQube Server. See Software requirements for more details.
PostgreSQL in Helm charts removed
The deprecated PostgreSQL dependency in the Helm chart has been removed. If you were relying on this dependency for production, you must take the following steps to upgrade to the new chart: back up their existing database, import the data into a new database, and then update the JDBC URL within the SonarQube chart configuration. See Installing Helm chart for more information.
Kubernetes and Openshift versions removed
Support for versions 1.30 and 1.31 has been removed.
Support for versions 4.11 to 4.16 has been removed.
2016 MSSQL Server 13.0 support removed
Support for 2016 MSSQL Server 13.0 support has been removed. See Installing database for more information.
Deprecation of Ingress NGINX
Due to the retirement of the ingress-nginx controller in November 2025 (with best-effort support ceasing in March 2026), the dependency on this chart is now deprecated.
We advise migrating to the Gateway API, which is the modern successor to Ingress. Should you need to continue using Ingress, consult the Kubernetes documentation for a list of suitable alternative controllers. A replacement dependency will be provided in a future release.
Deprecation of Automatic AI Code Detection
Autodetect AI-Generated Code has been deprecated. Sonar will adjust the AI Code Assurance offering to adapt to the industry changes with high AI adoption. A warning callout has been added to the SonarQube UI in global and project settings. See AI Code Assurance for more information.
Full release notes
Links to the full release notes in Jira:
Last updated
Was this helpful?