Lucee 8.0.0.294-RC available for testing

Lucee 8.0.0.294-RC is available for testing, the release candidate for Lucee 8.0.

What Lucee 8 is about

Stable, robust, fast. Lucee 8 is mostly about the engine your code runs on, not about new syntax. Lucee 7 started a lot of this work, and Lucee 8 finishes it:

  • Everything comes from Maven: core, extensions and Java libraries.
  • The configuration was rewritten: it is declarative, loads lazily, and every setting can be set from outside.
  • Performance work in every release.
  • Virtual threads all the way through.

Most of this makes Lucee faster and lighter without any changes to your code. The sections below go through each theme.

All in on Maven

Lucee 7 started moving to Maven. In Lucee 8, Maven is how Lucee gets everything it needs:

  • Extensions are Maven artifacts.
    • In .CFConfig.json (or LUCEE_EXTENSIONS) you can name an extension by its Maven coordinate, for example {"maven": "org.lucee:mysql-jdbc-extension"} for the latest version.
    • You can also point to a .lex file anywhere Lucee can read, with {"resource": "..."} (local, http, s3, …).
    • The old {"id": "..."} form still works.
    • Extension providers are now Maven group IDs (for example org.lucee) instead of URLs.
  • A smaller core.
    • Markdown and SMB moved out of core into extensions (LDEV-6520). Both are bundled, so markdownToHTML() and smb:// keep working.
    • Most bundled extensions are now lite builds: their Java libraries are fetched from Maven when the extension installs.
    • The Janino compiler is downloaded only when it’s first needed.
  • Classifiers everywhere. groupId:artifactId:version[:classifier][@extension] works in javaSettings, mavenLoad() and extension manifests (LDEV-6555).
  • More reliable installs:
    • fewer redundant requests and downloads (LDEV-6523)
    • the bundled artifacts are placed in the local Maven repository before extensions install (LDEV-6556)
    • extensions from LUCEE_EXTENSIONS without a version stay installed (LDEV-6591)
  • Your repository, your rules. Lucee uses Maven Central and maven.lucee-services.com by default, and you can point it at your own Maven-compatible repository or mirror with maven.repository.

Configuration, rewritten

We rewrote the configuration loading from scratch. It is now declarative: every setting is described once, with its type, default and documentation. Settings load when they are first used, and changing one setting no longer reloads the whole configuration. The main goal was speed, and we got more than that:

  • Cloud ready. Everything you can set in .CFConfig.json can also be set as an environment variable or system property. That includes mappings, datasources and everything else (for example LUCEE_COMPONENTMAPPINGS_1_VIRTUAL=...).
  • No silent overrides. A setting that comes from the environment wins over the file and shows as read-only in the Administrator. Trying to change it there throws an error instead of saving a value that would never be used (LDEV-6452).
  • Your editor knows Lucee’s config.
    • ConfigSchema() returns a JSON Schema for .CFConfig.json. Point VS Code or IntelliJ at it for validation, autocomplete and inline docs.
    • EnvVars() and GetSystemPropOrEnvVarInfo() list every system property and environment variable, with type, default, description and the config key each one maps to.
  • Consistent names. Administrator.cfc and cfadmin arguments now use the same names as the config keys.

Stable, robust, fast

In Lucee, performance isn’t a single ticket. It’s part of every release. In 8.0 the compiler, runtime and core types got lighter and quicker again:

  • Function calls:
  • Requests and memory:
  • Built for scale:
    • the application, server and session scopes use a concurrent map, which removes most lock contention under load (LDEV-6479)
    • less contention for CSRF tokens and thread names (LDEV-6517)
    • parallel iteration reuses pooled PageContext clones instead of creating one per element (LDEV-6370)
  • Java, faster and closer:
    • creating Java objects with createObject() / new is faster (LDEV-6614)
    • for-in and cfloop collection work with any Java Iterable or Collection (LDEV-6559)
    • SLF4J 2.x logging from Maven-loaded libraries ends up in the Lucee logs (LDEV-6664)
    • you can use a custom truststore for SSL certificates (LDEV-5571)
    • ASM 9.10.1 is ready for Java 27 (LDEV-6505)

Virtual threads

Lucee 8 requires Java 21, so virtual threads are always available. Virtual threads are lightweight threads managed by the JVM, so you get high concurrency without a platform thread for every task:

  • Parallel iterators. arrayEach(), arrayMap(), arrayFilter() and friends accept parallel="none" | "thread" | "virtual", plus maxConcurrency (LDEV-6369). parallel=true now runs on virtual threads.
  • <cfthread virtual="true"> runs a thread on a virtual thread, which is a good fit for I/O-heavy work. Set LUCEE_THREAD_VIRTUAL=true to make that the default for all cfthreads.
results = arrayMap( ids, function( id ) {
    http url="https://example.com/api?id=#id#" result="local.rsp";
    return deserializeJSON( rsp.fileContent );
}, "virtual" ); // none | thread | virtual

Tooling for your code

These two features came with 7.x and are part of the Lucee 8 toolset:

  • AST: astFromString() and astFromPath() turn CFML (tags or script) into a JSON syntax tree with exact source positions. You can build linters, formatters and other tools on top of it.
  • Compile to bytecode: compileToBytecode() compiles a file or directory to bytecode in place, and restoreToSource() brings the source back if you compiled with keepSource. You can also encrypt the bytecode with an RSA private key (generateRSAKeys()). The server then only needs the matching public key, set as PUBLIC_KEY, to run it.

Extensions

  • PDF 3.0 is bundled. <cfdocument> now renders with OpenHTMLToPDF, and <cfpdf> uses PDFBox 3.
  • Scheduling. The classic scheduler is still bundled. If you need more, the Quartz Scheduler extension (cron expressions, cluster support) and the Tasks event gateway extension are available as separate extensions.
  • Markdown and SMB are new bundled extensions (see above).

Language additions

  • the ?? null-coalescing operator (LDEV-5343)
  • an onMissingFunction() convention in Application.cfc for dynamic UDFs (LDEV-5094)
  • structIsOrdered() (LDEV-6345)
  • queryColumnNames() (LDEV-5748)
  • multi-character delimiters in listAppend() (LDEV-2024)
  • deserializeJSON( strictMapping="query" ) (LDEV-5349)
  • a precise argument for toBinary() (LDEV-6584)
  • inspectTemplate in getApplicationSettings() (LDEV-4931)
  • bracket notation for nested form and URL fields with formUrlAsStruct (LDEV-6070)

On the way to Lucee 9: regex

The perl regex engine (Apache ORO) is deprecated in 8.0 (LDEV-6660):

  • It is still the default, so nothing changes in behaviour, but Lucee logs a warning once per application that uses it.
  • Lucee 9 will default to the Java engine.
  • 8.0 adds a compat mode (LDEV-6432): the Java engine with Perl-style replacement syntax (\1, \U, \L), which makes the switch easier.

Try this.regex = { type: "java" }; (or "compat") on your applications now.

Upgrading from 7.1

Please read the breaking changes 7.1 → 8.0 before you upgrade. The most important points:

  • Java 21 is required. Lucee 8 won’t start on Java 11 or 17. Jakarta works as in 7.x.
  • Some .CFConfig.json keys must move into their section. The debugging settings go into monitoring, cacheDefault* into cache, and updateProxy* into proxy. Lucee 8 ignores them at the top level without a warning.
  • Environment variables win over .CFConfig.json for every setting, so check for leftover LUCEE_* variables.
  • parallel=true now uses virtual threads with no limit. If your closures use limited resources such as database connections, set parallel="thread" or maxConcurrency.
  • The application, server and session scopes no longer keep insertion order.
  • Maven coordinate strings: a non-scope value in 4th position is now read as a classifier.
  • PDF output: compare your important <cfdocument> output with 7.1.
  • Offline or firewalled servers need access to a Maven repository or a mirror (maven.repository), because lite extensions and Janino are resolved from Maven.

What’s new since 8.0.0.186-RC

  • All bundled extensions are final releases now. Markdown and SMB are bundled for the first time.
  • The perl regex deprecation, Maven classifiers, and the scope and parallel changes described above all arrived in this RC.
  • The fixes from the 7.1 line are merged in, among them time zones, null handling, mail, datasource pools, sessions and the Administrator.
  • About 190 bug fixes and 17 compatibility fixes in total. The full lists are below.

Bundled extensions

Extension Version
MySQL 26.7.0
Microsoft SQL Server 13.4.0.jre11
PostgreSQL 42.7.11
Administrator 1.0.0.9
Documentation 1.0.0.8
S3 2.1.0.15
PDF 3.0.0.2
Image 3.1.0.11
ESAPI 3.1.0.7
Scheduler Classic 1.0.0.2
Compress 2.1.0.7
Mail 1.1.0.17
FTP 1.0.0.9
Markdown 1.0.0.2 (new in 8.0)
SMB 1.0.0.10 (new in 8.0)
Compatibility fixes since 8.0.0.186-RC

These make Lucee behave like other CFML engines. Please check them against your code.

  • LDEV-2122 IsValid(‘email’, ‘email@example.com,’) inconsistency with ACF
  • LDEV-2417 Fat arrow functions shouldn’t require parentheses around single parameter
  • LDEV-2454 ACF Compatibility for wildcard in this.blockedExtForFileUpload
  • LDEV-2672 left operand of the Elvis operator has to be a variable or a function
  • LDEV-2924 query.columnList cannot be used with list or string member functions
  • LDEV-2970 arrayslice() negative offsets off by one
  • LDEV-3188 Incompatibility - when using directoryList() with listinfo=“name”
  • LDEV-3739 site-wide error page doesn’t have “error” variable like ACF
  • LDEV-4062 Arrow function with no body errors in Lucee, but not in Adobe
  • LDEV-4122 Incompatibility - Multiple arguments with the same name in UDF doesn’t throw an error
  • LDEV-4160 Chaining method on returned object from returned UDF gives syntax error
  • LDEV-4250 script “break” requires semicolon in Lucee, but not Adobe
  • LDEV-4495 Struct.find should support a callback as the first arguments, as per CF
  • LDEV-4792 Can’t call lambda that exists in static scope
  • LDEV-5803 cross engine listDeleteAt function incompatibility
  • LDEV-5857 ACF Incompatibility: ReplaceNoCase() Throws Error When Substring Is Empty
  • LDEV-5890 Regression: can’t parse dates string produced using (long, medium, short) masks

The full list of bug fixes since 8.0.0.186-RC is in the first reply below.

The complete list of changes is in the GitHub compare.

Please test

This is a release candidate, so please try it with your applications and tell us how it went here or in Jira. Feedback on upgrading from 7.1 is especially welcome, and so is anything the breaking-changes doc doesn’t cover yet.

3 Likes

All bug fixes since 8.0.0.186-RC:

All bug fixes since 8.0.0.186-RC
  • LDEV-69 param does not support associative array syntax
  • LDEV-285 Lucee try to parse coldfusion tag in commented line.
  • LDEV-296 Throw an error on unsupported component initMethod
  • LDEV-935 Cfbreak always run after cfcontinue with cfsilent
  • LDEV-1818 Interface Methods require semicolons
  • LDEV-1962 Generated accessors keep references to old CFC when mixed into another CFC’s variable’s scope
  • LDEV-2027 QueryFilter, QueryMap, QueryReduce changes currentRow inside a loop
  • LDEV-2133 cfabort thows an error in cfscript implementation with type=“page”
  • LDEV-2213 Cannot set calling scope vars from closures when localMode=“modern”
  • LDEV-2242 Implicit getters strong typed in metadata to string
  • LDEV-2701 Lucee preceision issue
  • LDEV-2734 Tag island inside of tag-based try causes parsing error
  • LDEV-2773 unscoped thread variable not being found
  • LDEV-2899 booleanFormat(null) throws Can’t cast String to a boolean
  • LDEV-3020 Invalid Bytecode: Expecting a stackmap frame at branch target
  • LDEV-3046 isValid(“email”,“test example@exaaple.com aaa”) returns true
  • LDEV-3095 isValid(“email”) fails on examples from wikipedia
  • LDEV-3213 local class incompatible: stream classdesc serialVersionUID
  • LDEV-3335 Store component properties and accessor UDFs statically per class
  • LDEV-3393 For-in loop throws NullPointerException when located within finally block
  • LDEV-3619 param this.that throws error if preceded by an extra space
  • LDEV-3671 function metadata does not handle final arguments
  • LDEV-3672 array out of bounds when dump CFC object with final property
  • LDEV-3847 CFQUERY using attributeCollection fails when inside CFFINALLY
  • LDEV-3873 Array.every incompatibility with ACF
  • LDEV-3914 thread statement cannot be used inside an arrow function
  • LDEV-3963 Colons (:) in function metadata is treated as and equal sign (=) instead of part of the identifier
  • LDEV-4021 Throwing serialized/deserialized exception object loses original stacktrace and tag context
  • LDEV-4023 arrayEach/Reduce ignores elements with null values
  • LDEV-4151 concatenation operator fails inside cffinally inside a function
  • LDEV-4157 Missing parent Statement of Statement error with thread and tag island
  • LDEV-4211 Using not existing ORM entity within transaction throws incorrect error
  • LDEV-4246 Static methods fail when called from an abstract class
  • LDEV-4268 CFML parsing bug with continue
  • LDEV-4286 nested function(){}; throws byteccode error
  • LDEV-4366 cfparam takes any random attribute as default value
  • LDEV-4424 Query Params are not Resolved When Query has Question Mark
  • LDEV-4665 Query filter/map functions alter datatypes
  • LDEV-4685 cfloop condition throws error when in tag island
  • LDEV-4698 QoQ SQL parser doesn’t understand cast with lengths
  • LDEV-4709 isClosure Returns false on Arrow Functions
  • LDEV-4771 function scope context cascading and cfloop query
  • LDEV-4776 decimalFormat wrong rounding
  • LDEV-5027 Administrator.cfc does not allow empty default in updateCacheConnection()
  • LDEV-5143 CFDBInfo Fails on PostgreSQL Tables Created with Quoted Names
  • LDEV-5193 Caster methods like toShort, toInteger, toByte don’t check for truncation or overflow
  • LDEV-5212 executionLog misses function name when calling from object
  • LDEV-5314 Time Zone Id issue (IST) in 7.0.
  • LDEV-5352 compiler error: Type lucee/runtime/type/Closure not present
  • LDEV-5360 Query Indexed not working with QueryOfQuery
  • LDEV-5461 inspectTemplates() and pagePoolClear() should call ApplicationPathCacheClear()
  • LDEV-5486 LSEuroCurrencyFormat() with Empty Locale Showing Illegal Sign
  • LDEV-5519 wrong constructor called on org.json.JSONObject
  • LDEV-5533 NPE on invalid inline component syntax
  • LDEV-5547 Property Default Values Type Inconsistency
  • LDEV-5698 admin: extension list labels lack contrast in dark mode
  • LDEV-5713 CVE-2025-48924: Apache Commons Lang
  • LDEV-5726 FileExists() Returns False When URL Contains Spaces
  • LDEV-5737 expired one off scheduler tasks shouldn’t start a thread and should stop after execution
  • LDEV-5747 concurrent cfschedule updates causing CFConfig corruption
  • LDEV-5768 Admin: Dark Mode Should Automatically Follow System Settings (Lucee 7)
  • LDEV-5786 Admin (Dark Mode): Improve Text Contrast and Update version Button Color Differentiation in Lucee 7
  • LDEV-5790 QoQ incorrect results with large numbers
  • LDEV-5807 Regression with java objects running spreadsheet-cfml
  • LDEV-5834 regression: cfexecute listeners corrupt scopes
  • LDEV-5837 extension version check error
  • LDEV-5841 expandPath behaves strangely with JS223
  • LDEV-5852 Admin: “Cancel” button on Custom Tags list page resets all form checkbox selections globally
  • LDEV-5912 Filecopy with HTTP ResourceProvider (aka http resource) makes 5 HEAD requests before downloading
  • LDEV-5922 Operand stack underflow when using for-in loop with try-catch inside finally block
  • LDEV-5932 CFHTTP GET doesn’t enable compression due to HttpGetWithBody
  • LDEV-5952 internalRequest fails with “Content was already flushed” after resetBuffer()
  • LDEV-5966 Request timeout causes phantom connections in datasource pool
  • LDEV-5969 AST does not capture cffunction/cfargument tag attributes
  • LDEV-5975 AST dump loses method name in chained method calls
  • LDEV-5977 AST Ternary expression alternate value duplicates consequent
  • LDEV-5978 AST queryExecute adds extra variable name argument
  • LDEV-5979 AST isDefined adds extra scope argument
  • LDEV-5980 AST BIF calls keep adding internal metadata
  • LDEV-5982 AST cflock adds internal id attribute
  • LDEV-5983 AST cffunction body duplicates cfargument comments
  • LDEV-5984 AST Escaped hashes lose escaping
  • LDEV-5985 type=“java” functions crash AST parser
  • LDEV-5986 AST Tag islands converted to strings
  • LDEV-5987 AST Missing parent statement error
  • LDEV-5989 AST Interpolated attribute values have escaped hashes in raw field
  • LDEV-5990 AST Function param hint/defaultValue metadata lost
  • LDEV-6006 astFromString() breaks on CFML comment syntax in string literals
  • LDEV-6007 AST Component inside cfscript becomes sibling instead of child
  • LDEV-6008 PreserveSingleQuotes evaluated in cfquery AST
  • LDEV-6011 Internal function names in AST instead of literal syntax
  • LDEV-6013 AST MemberExpression with string key has computed=false
  • LDEV-6014 AST Dynamic property assignment with interpolated key loses structure
  • LDEV-6016 Static blocks lose static marker in AST
  • LDEV-6017 AST Interpolated struct keys become BinaryExpression CONCAT
  • LDEV-6018 AST String concatenation wraps parts in CastExpression with typeAnnotation
  • LDEV-6019 Unquoted function attributes after parenthesis fail AST generation
  • LDEV-6022 AST String literal concatenation merged at parse time
  • LDEV-6024 Computed member expression inside scope access corrupts AST
  • LDEV-6027 Property attributes reordered in AST
  • LDEV-6028 Struct key AST type changes based on separator (: vs =)
  • LDEV-6029 AST Script function cachedWithin pre-evaluates timespan expressions
  • LDEV-6030 AST Labeled loops lose their labels
  • LDEV-6031 AST cfimport, leaks internal attributes, omits name
  • LDEV-6032 AST Boolean attribute values parsed incorrectly in both tag and script modes
  • LDEV-6034 AST should not expose auto-generated closure/lambda names
  • LDEV-6035 AST lacks fullname for shorthand script tags like abort; vs cfabort;
  • LDEV-6036 AST adds default type=“any” to property when not in source
  • LDEV-6038 Comments become empty StringLiteral placeholders in AST
  • LDEV-6041 AST should include original source fidelity
  • LDEV-6078 Bytecode has invalid line numbers, breaking step debugger and creating misleading exceptions
  • LDEV-6119 ClassicLayout pattern merges message and stacktrace w/out space
  • LDEV-6122 XMLNodeStruct.equals() mutates the other object (assignment instead of comparison)
  • LDEV-6145 OSGiUtil.improveFileName() renames bundle JAR without updating Felix, causing “not unique” crash on restart after extension upgrade
  • LDEV-6154 QoQ string concatenation with || operator excludes NULL and empty string rows
  • LDEV-6166 StructImpl.TYPE_MAX (AccessOrderLimitedSizeMap) is not synchronized
  • LDEV-6188 Behavior change — optional arguments absent from arguments scope when nullSupport=true
  • LDEV-6202 QueryExecute passes entire options struct as tag attributes, leaking datasource keys
  • LDEV-6218 cacheRemove with cachedWithinId does not work for HTTP cache entries
  • LDEV-6233 QueryImpl: currentRow state survives serialisation, causes stale row reads
  • LDEV-6236 Optimise generated accessor UDFs — skip duplication and bypass UDF dispatch
  • LDEV-6302 toSorted() required a sort direction
  • LDEV-6303 cfproperty default=“#expr#” no longer surfaces via Property.getDefault()
  • LDEV-6305 cfqueryparam CF_SQL_FLOAT against MSSQL FLOAT column returns no rows
  • LDEV-6313 OSGiUtil._loadBundle drops cause when bundle download fails
  • LDEV-6350 SpoolerEngineImpl not cleaned up during engine restart
  • LDEV-6354 OSGiUtil.toVersion(String,Version) throws on malformed qualifier, aborting S3 BundleProvider listing (e.g. “javassist-3.9.0.GA copy.jar”)
  • LDEV-6447 sessionInvalidate() creates replacement sessions and can throw after response commit
  • LDEV-6449 getDisplayPath() can return null, causing NullPointerException in debugger
  • LDEV-6455 Spooled cfmail silently dropped - NotSerializableException: sun.nio.cs.UTF_8 when persisting spooler task (Lucee 7.1 / Mail Extension 1.1.0.6)
  • LDEV-6458 Mail admin UI / SMTP server configuration missing after mail moved to extension (7.1.1.x)
  • LDEV-6462 Removing Classic Scheduler extension also removes /lucee/admin mapping
  • LDEV-6469 Regression: GetMetadata( childCfc ).functions no longer includes inherited methods
  • LDEV-6472 lucee-light warmup skips Server.cfc onBuild() on first pass (needs second start)
  • LDEV-6475 Java serialisation: Array serialVersionUID breaks Lucee 7 deserialise of Lucee 8 streams
  • LDEV-6476 Java serialisation: Soft struct cannot round-trip serialize/deserialize in Lucee 8.0
  • LDEV-6481 NPE on clean boot when Event Gateways auto-start before Component.cfc is deployed
  • LDEV-6482 Inspect Templates Auto stops detecting CFM/CFC changes after LDEV-6358
  • LDEV-6483 Server.cfc (onServerStart/onBuild) silently not executed when server .CFConfig.json has no mappings beyond /lucee and /lucee-server
  • LDEV-6484 new mail(…).send() fails on 7.1 with “not supported in 7.1 OOTB” (HelperBase) — regression, restore via mail extension
  • LDEV-6485 cfmail fails JVM-wide with “jakarta.mail.Provider: com.sun.mail.imap.IMAPProvider not a subtype” once javax.mail was used (7.1 / 8.0)
  • LDEV-6486 Minor update from 7.0 to 7.1 (.lco or Admin update) does not install the Mail extension: its Require-Extension entry has no “since”
  • LDEV-6487 isValid(“email”) on 7.1 / 8.0 uses a temporary regex: rejects o’reilly@example.com, accepts .foo@, user.@ and a…b@ (regression)
  • LDEV-6489 Spooler logs java.io.EOFException for an empty or truncated .tsk file, deletes the file and does not say which file it was
  • LDEV-6490 getMetadata() throws “this.cpRef is null” on a component restored with objectLoad(objectSave()) (7.0 regression)
  • LDEV-6492 ExtensionProvider.detail() overwrites lite .lex URL with lite.asc GPG signature for SNAPSHOT classifiers
  • LDEV-6503 Request timeout returns a connection with an unread result to the pool, the next request gets another query’s result (regression since 6.2.5.6)
  • LDEV-6506 SystemUtil.JAVA_VERSION is 0 on GA Java releases (e.g. 21, 25, 27)
  • LDEV-6507 Remove Connection: close on cflocation, REST error responses and cfflush interval
  • LDEV-6508 DecimalFormat number to string conversion serialises on a global lock
  • LDEV-6510 continue inside a try/catch leaves the loop like a break when the same try/catch also contains a break
  • LDEV-6512 getInspectTemplate() reads sysprop / env var on every call (8.0)
  • LDEV-6515 Admin access setting “open” is ignored by cfadmin since multi-mode removal: the server context always requires a password
  • LDEV-6523 Maven extension resolver: avoid redundant requests and re-downloads (.pom/.lco probes, HEAD before every GET, install moves the .lex out of the local Maven repo; 8.0 fetches the .lex as .pom)
  • LDEV-6533 final in a component pseudo-constructor leaks a DataMember object into the variables scope
  • LDEV-6543 QoQ: cast() inside a sub-query fails
  • LDEV-6544 Function listener / cfthread separateScopes=false: caller’s local and arguments scopes are recycled while used
  • LDEV-6545 CFMAIL: HTML parts are sent 7bit again (LDEV-5176 flipped the quoted-printable default), long lines break
  • LDEV-6546 Cache tag clear can abort with NullPointerException and leave stale entries; RamCache tag clear deep-copies every entry
  • LDEV-6547 dateFormat/timeFormat with an X mask changes the JVM default timezone to CET
  • LDEV-6548 Server proxy from .CFConfig.json is ignored in Lucee 7 (getProxyData() always null)
  • LDEV-6553 Datasource names with dots are rejected by the Administrator
  • LDEV-6562 Datasource connection pools leak after idle eviction (POOL_MAX_IDLE): pools not closed, MBeans stay registered
  • LDEV-6567 Datasource pool creation can leave a replaced pool unclosed (use putIfAbsent and close loser)
  • LDEV-6570 Time values (createTime / createODBCTime / cfschedule startTime) are shown 1 hour early in America/Vancouver with timezone data 2026b
  • LDEV-6573 Class::init() can’t call a Java static method named init
  • LDEV-6575 invokeImplicitAccessor: implicit getter missing for property in sub component when called inside parent file
  • LDEV-6579 Maven downloads should use the common downloader
  • LDEV-6581 Internal downloads should use the server proxy - 8.0
  • LDEV-6586 Config placeholder replacement turns ordered structs into unordered ones
  • LDEV-6587 createDynamicProxy / implementsJava: returning this or a CFML array to Java throws ClassCastException
  • LDEV-6588 Closures / UDFs as Java interfaces fail with VerifyError
  • LDEV-6591 Lucee 8: extensions installed via lucee.extensions / LUCEE_EXTENSIONS without a version are uninstalled again at start-bundles
  • LDEV-6594 CSRF token lost after structClear(session) with storage-backed sessions and sessionCluster=true
  • LDEV-6596 Session writes and CSRF tokens lost with storage-backed sessions (Redis) and sessionCluster=true
  • LDEV-6598 Logging uses the request/application timezone or the JVM default instead of the server config timezone
  • LDEV-6610 Typed argument with null default fails when omitted with full null support
  • LDEV-6612 Lucee 8: maven coordinate without version resolves the latest Central release instead of the locally installed version
  • LDEV-6616 Admin: extension names in the extension tiles are too large for the fixed-size tile
  • LDEV-6617 Admin: restart page uses a fixed 740px wide table
  • LDEV-6639 mavenLoad resolves optional (true) transitive Maven dependencies
  • LDEV-6665 Lucee Light logs spurious ERROR “Could not find [/extensions/.index]” on clean boot
  • LDEV-6666 Fresh install: “could not load the base component” errors because extensions install before core base CFCs are extracted
  • LDEV-6671 Maven download policy error message names non-existent property lucee.maven.download.policy
  • LDEV-6673 resetLoggers() drops the logger map without closing the appenders, so every log gets a second appender after a config load
  • LDEV-6680 S3 extension: exists() reports missing and caches NotExisting on connection/pool errors
  • LDEV-6699 Guard 3.x: decodeFromURL() throws on a % that is not a valid escape
  • LDEV-6712 Light and zero jars still contain the bundled maven/ folder
  • LDEV-6713 Lucee 8: admin action=“removeRHExtension” throws a NullPointerException for an extension that is not available from any provider
1 Like

While I was testing the 7.1.2 release I just tried 8.x as well and same results - all my tests are passing!

1 Like

There were two breaking changes for me, but I was able to resolve both:

  1. Removed "psq": "false", line from dataSources in .CFConfig.json

  2. Modified app code to work with new support for form input bracket notation (LDEV-6070).

Thanks Jim and Kenric for testing, great to hear!

Kenric, the psq setting inside dataSources is a bug, not an intended change. It’s fixed under LDEV-6751 and will be in the next 8.0 build, so you can keep your config as it was.

The bracket notation (LDEV-6070) is intentional: user[name]=x now becomes form.user.name, the same way jQuery, PHP and Rails handle it. If an app relies on the old literal keys, you can turn it off with LUCEE_FORMURLASSTRUCT_BRACKET_NOTATION=false (or -Dlucee.formUrlAsStruct.bracket.notation=false). We’ll add it to the 8.0 breaking changes page.

Will 8.0.0.x Release happen around the same time as the 7.1.2.x Release?

I’m just wondering if I should skip right ahead to Lucee 8.

Hi Kenric, we now let every RC soak for 10 days before the final release, and any reported regression restarts that cycle. If all goes well, 7.1.2 will be released around 5 November. The psq issue you reported is fixed under LDEV-6751 and will be in the next 8.0 RC, so 8.0 final will follow about 10 days after that RC, assuming no further regressions.

Which one to pick depends on what you need. 7.1 will soon become the LTS version, so it gets bug and security fixes only, with no functional changes, and it’s the safe choice for production. 8.0 will then be the cutting-edge release with the newest features. Moving straight to 8 is fine if you’re comfortable with that, and testing the RCs now really helps us. Either way, let us know if anything comes up.

1 Like

I decided to take the plunge and updated to 8.0.0.294-RC along with the latest Guard, Mail, and FTP Releases.

In production about 48 hours now and zero issues.

2 Likes

Lucee 8.0.0.299-RC (RC2) is available for testing. It is a small update on top of 8.0.0.294-RC: a few fixes for problems found while testing RC1, and the bundled Mail extension is updated.

Changes since 8.0.0.294-RC

  • Extensions: installing or updating an extension that contains TLDs no longer fails with a NullPointerException (LDEV-6720).
  • Extensions: fixed a deadlock on Java 21 while the extensions were loading (LDEV-6812).
  • Core upgrade: the compiled classes (cfclasses) are purged again when you upgrade the core (LDEV-6723).
  • Configuration:
    • a legacy psq key inside dataSources in .CFConfig.json no longer breaks every datasource (LDEV-6751).
    • the server-level nullSupport setting is honoured for argument defaults when there is no Application.cfc (LDEV-6748).
    • dotNotationUpperCase in .CFConfig.json is honoured again (LDEV-6749).
    • checkForChanges reloads the config file again after it was changed outside Lucee (LDEV-6813).
  • Datasources: “Verify” in the Administrator uses the JDBC driver of the extension, which fixes the MSSQL datasource creation error (LDEV-6443).
  • Mail: the bundled Mail extension is now 1.1.0.24, which fixes the NoClassDefFoundError for org/apache/commons/text/WordUtils (LDEV-6593).

Thanks to everyone who tested RC1 and reported these.

Please test

Please try this RC with your applications and tell us how it went, here or in Jira. If you hit a problem on RC1, it’s worth checking whether it’s one of the above.

This RC soaks until Friday, 6 November 2026. If no regressions turn up by then, the final Lucee 8.0 release is planned for that day. A reported regression moves the date.

All tests passed again.

1 Like

I am back with another report from Claude:


8.0 regression: name[] form/URL fields no longer arrive as arrays (LDEV-6070)

Affects: 8.0.0.260 and later

Repro

test.cfm:

<cfoutput>#isArray( url.tags )# #serializeJSON( url.tags )#</cfoutput>

Request test.cfm?tags[]=lucee:

  • 6.2 / 7.1: true ["lucee"]
  • 8.0: false "lucee" (and tags[]=java&tags[]=cfml gives "java,cfml")

Same in the form scope, and for nested names: user[tags][]=foo&user[tags][]=bar and
user.tags[]=foo produce a nested struct whose tags is a string instead of an array.

This is the shape jQuery’s $.param/$.ajax sends for arrays (traditional: false,
the default), and what URLSearchParams.append( 'tags[]', … ) callers send, so any
param name="url.tags" type="array" or isArray( form.tags ) behind those calls now fails.

Cause

LDEV-6070 (ada529851, 8.0.0.260) routes every name containing [ through the new
ScopeSupport.parseFormUrlName(). That parser skips empty bracket segments ("skip empty segments like []"), so tags[] comes out as the single segment tags, and the
name.endsWith("[]") check in _fill() that creates the array never sees the marker. A
second value is then comma-joined onto the first.

The LDEV6070 test has a describe block "Array notation with [] suffix (existing functionality)", but its test.cfm prints arrays with arrayToList(), so the expectation
tags=java,cfml,lucee is satisfied by both the array and the comma string, and the
regression is invisible to it.

lucee.formUrlAsStruct.bracket.notation=false isn’t a workaround: it brings tags[] back
but makes user.tags[] a flat key literally named user.tags, with no nested struct,
which 6.2 and 7.1 both produce.

Suggested fix

Keep a trailing [] on the previous segment, so _fill() creates the array exactly as
before. Other empty segments (user[]name) are still skipped, so every existing LDEV6070
expectation is unchanged.

--- a/core/src/main/java/lucee/runtime/type/scope/ScopeSupport.java
+++ b/core/src/main/java/lucee/runtime/type/scope/ScopeSupport.java
@@ -252,10 +253,15 @@ public abstract class ScopeSupport extends StructImpl implements Scope {
 					// Closing bracket without opening - malformed
 					return null;
 				}
-				// Save segment inside brackets (skip empty segments like [])
 				if (i > start) {
 					segments.add(name.substring(start, i));
 				}
+				// A trailing [] is the array marker (tags[], user[tags][]): keep it on the
+				// previous segment so _fill() creates an array; other empty segments are skipped
+				else if (i == len - 1 && !segments.isEmpty()) {
+					int last = segments.size() - 1;
+					segments.set(last, segments.get(last) + "[]");
+				}
 				inBracket = false;
 				start = i + 1;
 			}

I believe that breaking change is a feature. Workarounds to restore previous behavior are also mentioned in LDEV-6070, although it appears that Claude also gave you at least one solution (there are probably others) if you have a budget for modernizing your Lucee code.

Lucee 8.0.0.300-RC (RC3) is available for testing. It is a small update on top of 8.0.0.299-RC: a fix for a regression reported while testing RC2.

Changes since 8.0.0.299-RC

  • URL and form scopes: name[] fields (for example tags[]=red&tags[]=blue) arrive as arrays again, as they did in 6.2 and 7.1. LDEV-6070 had dropped the trailing [] array marker, so those values became a plain string (LDEV-6889, reported by John Berquist). This was a regression, not an intended breaking change.

Thanks to everyone who tested RC2 and reported this.

Please test

Please try this RC with your applications and tell us how it went, here or in Jira. If you hit the name[] issue on RC2, this build should restore the previous array behaviour.

Because this was a reported regression, the soak period starts again from this RC. This RC soaks until Monday, 16 November 2026. If no regressions turn up by then, the final Lucee 8.0 release is planned for that day. A reported regression moves the date.

Thanks John, great report, and your diagnosis was right. LDEV-6070 turned tags[] into a plain string because the parser skipped empty [] segments, so _fill() never saw the array marker. The fix is under LDEV-6889: a trailing [] is kept on the previous segment, exactly as you suggested.

This is in 8.0.0.300-RC (RC3), available now:

This was a regression, not an intended breaking change, so the soak period starts again from this RC. Please give it a try when you have a chance.

i decided to revert this, because in that case for me backward compatibility is more important.

1 Like

The description of LDEV-6070 specifically states “Breaking Change”.

Well then I’m surprised that all my tests are still passing with 8.0.0.300-RC because I had already updated my app to work with the new bracket notation paradigm.

Hi Kenric, that matches what RC3 does. The fix only restores the trailing [] array marker, so tags[]=a&tags[]=b is an array again as in 6.2 and 7.1. The nested bracket notation from LDEV-6070 (for example user[name]=x) is unchanged, so an app already updated to that style keeps working on 8.0.0.300-RC. If you see anything that doesn’t behave like that, please tell us what you’re sending.