Showing results 526 to 550 of 1561

Feature
New Orphaned Backend User Interface

#99422 Show/hide unused content in the page module only if the user check/uncheck the "show unused content" checkbox

Hello,

We want the editors stop hidding useless contents instead of deleting them, and cluttering their own working area.

This improvement is splitted in smaller, and if possible, independant patches.

link to the full video (publically accessible): https://drive.google.com/file/d/1cxHaU7JHPiDE1_4hFsfDiM1m2YxfHoLz/view?usp=share_link

This tasks relates to the time 1.20min of the video

Rachel

Author Rachel Foucard
Created 12/23/22, 16:27
Updated 01/10/23, 16:17
Bug
New Orphaned Site Handling, Site Sets & Routing TYPO3 CMS 11

#90736 Wrong slug fields for siteroots (level > 1) in multidomain environment and no possibility to modify them

In a multidomain environment (where the sites are not located in the first level) the “is_siteroot” pages slugs are generated wrong. In addition, there is currently no possibility to manage them without directly modifying the slug in the database.

For example, if a new page is created in level 2, depending on the way it is created (wizard and from type of the parent from which executed) the slug field is shown in tca form. Even if slug is set to “/” here, “/1” or similar is saved instead. After that the field is not shown ony more at all.

Suggestion: At least when the created page is set to “is_siteroot”, the slug should automatically bet set to “/”.

Regarding multi language sites (like in my case), the behaviour should be checked, too. But it looks like the translated pages slug ist correctly set to “/”.

I am using version 9.5.14.

- Root [0]
-- Start [1] (Just for assigning some global TS, often use case)
--- Domain A [is_siteroot + site configuration] -> Slug is set to "domaina" or (if set to "/" if form is shown) to "/1"
--- Domain B [is_siteroos + site configuration] -> Slug is set to "domainb" or (if set to "/" if form is shown) to "/2"
Author Jan Kornblum
Created 03/11/20, 17:45
Updated 11/01/22, 17:24
Bug
New Orphaned TYPO3 CMS 11

#98457 Processed file doesn't exist

Use case: configure TYPO3 so that f:uri.image generated files are created inside typo3temp instead of fileadmin folder.

Steps to reproduce:
1. have ‘fileadmin/ (auto-created)’ storage with uid=1
2. have ‘typo3temp’ storage with uid=2
3. setting processingfolder field of ‘fileadmin/ (auto-created)’ storage to ’2:assets/_processed_’

This setup results in f:uri.image renders the correct path (eg: /typo3temp/assets/_processed_/e/6/959161_10151652466566963_1259477499_o_5a95879fc3.jpg) but the file doesn’t exist.

Author Christian Toffolo
Created 09/28/22, 09:30
Updated 09/28/22, 09:30
Bug
New Orphaned Backend User Interface TYPO3 CMS 11

#92664 Some input elements are cut off when resized and cause horizontal scrolling

A user should be able to resize a website up to 200% without text getting clipped or truncated. It mostly works when the user hides the .scaffold-content-navigation.

There are some input elements that are cut off:

  • In General: URL Segment
  • In Behaviour: Link Target
  • In Resources: Include static Page TSconfig (from extensions)
  • In Resources (Lightbox File selector): “Upload files”, “Add new media asset” and “Create new folder”
  • In Access: Usergroup Access Rights
  • In Social Media: Type of card to show
  • The date picker

#tas2020

Author Jakob Widhalm
Created 10/22/20, 10:42
Updated 09/24/22, 14:42
Bug
New Orphaned System/Bootstrap/Configuration TYPO3 CMS 10

#98124 potential dangerous destructor order

TLDR: classSchemata property of ReflectionService is saved to data cache (extbase) inside the destructor the class. But the destructor of the used cache frontend is called before the destructor of ReflectionService. Currently this isn’t a problem, because the standard cache frontends doesn’t contain a destructor.

I’m currently writing an extension to collect live statistics of cache usage in typo3. For this i replace the standard frontends with other frontends which contains some counters. The counters must be persisted between two calls, so i write the data to database inside the destructor of my frontends. In the stored statistics i noticed there must be some missed cache writes. Some investigations showed this is related to how the reflection service stores his data to the cache. And i found out that the destructor of my replaced frontend class was called before the destructor of the reflection service. I’m not sure if my replaced frontends cause a change in the destructor order and if the destructor order would be the same on all installations using the same configuration.

As said earlier, currently this isn’t causing problems in typo3, but it could if there would be a cache frontend which closes the connection to the cache backend in it’s destructor.

Solution: The reflection service should write his cache data before the first destructor is called. Potential problem could be, if there is a class which uses the reflection service inside of its destructor.

Author Mordamir
Created 08/11/22, 16:57
Updated 08/11/22, 16:57
Bug
New Orphaned Localization TYPO3 CMS 9

#84682 TYPO3 8 Relationship Localization Bug

!When it comes to bidirectional relationships, the localization handling of TYPO3 8 is a bit different than I expected to be.

Let’s say, we use price tables and seasons which are connected through a “has many relationship”.
I then create a new season and assign them to a price table.
Whenever I translate this season, its connections will also be translated because a new record will be inserted in the mm table.
If I then edit the assigned price table, the season appears twice ( original + language record ).
Additionaly, the sorting attribute of the mm table is broken because the language record receives the default value of sorting ( 0 ).

*I’ve created a minimalistic extension to demonstrate this behaviour.*
[[https://github.com/sir-script/mm_typo3_8]]

Can anybody explain,what’s wrong here?
What would you expect as standard behaviour?
Should the mm table contain these language overlay records and the presentation is wrong?

Author Michael Rainer
Created 04/10/18, 09:47
Updated 08/05/22, 07:24
Feature
New Orphaned Extbase

#97323 Extbase queries should respect the pages.hidden field when respecting storage pages

Currently Extbase only has a table.pid IN (...) condition-style when respecting storage pages. But actually it should do a join on pages and also query on pages.hidden = 0. This would be consistent to other core behavior (if I disable a page in Backend then also it’s content is implicitly disabled). At least there should be a way to optionally enable this behaviour, if not being default.

Author S P
Created 04/07/22, 05:45
Updated 04/07/22, 05:47
Feature
New Orphaned Backend User Interface

#97278 Add multiple selection of records in IRRE as well

Add the multiple selection used in the List Module also to the records in IRRE as well so as to allow faster bulk operations such as deleting a large number of IRRE elements.

Author Christian Toffolo
Created 03/31/22, 18:17
Updated 03/31/22, 18:17
Task
New Orphaned TYPO3 CMS 11

#97002 Simplify flexform behavior and stop saving old settings in flexforms (+ get rid of the then no more necessary flexform cli command + scheduler task)

When working with the forms framework I did come over some pretty nasty core behavior in the flexform context.

I did find parts in the flexforms like:

(...)
        <sheet index="ea4ca6637d2b7dbe41c867c14d43ef95">
            <language index="lDEF">
                <field index="settings.finishers.Redirect.pageUid">
                    <value index="vDEF">59</value>
                </field>
                <field index="settings.finishers.Redirect.additionalParameters">
                    <value index="vDEF"></value>
                </field>
            </language>
        </sheet>
        <sheet index="9cc4c6a803d8627cf0c4aca3161f4314">
            <language index="lDEF">
                <field index="settings.finishers.Redirect.pageUid">
                    <value index="vDEF">59</value>
                </field>
                <field index="settings.finishers.Redirect.additionalParameters">
                    <value index="vDEF"></value>
                </field>
            </language>
        </sheet>
(...)

So duplicated settings values just because the core flexform behavior doesn’t get rid of old settings when saving a flexform. (in this case the “persistenceIdentifier” changed)

I asked myself what is the purpose of saving old settings to the flexform fields? Why not only saving the current state? I don’t see a valid use case for keeping old settings values. (please switching plugins and restoring settings is NOT a use case)

In the end simplifying this would also make it possible to get rid of the “cleanup:flexforms” cli job. Which is just another maintenance task that should not be necessary.

Author Denis Mir
Created 02/23/22, 07:52
Updated 02/23/22, 07:52
Feature
Accepted Orphaned Install Tool

#96652 Reduce amount of weak matches in Extension Scanner

Scenario

Just as an example, analyzing ext:bootstrap_package in the Extension Scanner results in a bunch of weak matches concerning calling method ->error – which happens since TYPO3\CMS\Core\TypoScript\Parser\TypoScriptParser->error is configured to be reported. Attached screenshot shows a bunch of false-positives (for $this instead of expected TypoScriptParser).

The scope of this issue is to reduce the amount of weak matches that do not consider the proper instance type when invoking methods.

Investigation in using PhpStan

It seems to be possible to resolve those symbols and their type with PhpStan (using a custom rule and basically their Scope object), however there are some aspects that make it difficult for the Extension Scanner:

  • all used/called classes need to be available or to be stubbed
  • this requirement raises memory consumption (had to raise limits to 96M for a single file to be analyzed)
  • invoking standalone scans another PHP class seems to be difficult (but somehow still possible)

Investigation in extending current custom PHP Parser

In order to keep track of types that are exposed in a particular variable (or class property, or array-dim, or return-type of another function/method call, …) an own call-graph and scope needs to be maintained. It is just relevant for this feature to fetch the class-type (if any), not the actual value to keep memory consumption low.

In particular the Extension Scanner would first have to scan all symbols, maintain the call-graph and scope and then start analyzing symbols and compare them with corresponding configuration files. That’s some work…

General question

Concerning potential implications of using another tool, the main question is, whether the Extension Scanner shall stay accessible via Install Tool in browser, or whether that component is supposed to be a development tool that is invoked via CLI only.

—-

Author Oliver Hader
Created 01/26/22, 11:09
Updated 01/27/22, 13:11
Bug
New Orphaned TYPO3 CMS 10

#96290 ActionController->forward does not work when targetting another extension

I have a created some extensions for our instance of Typo3.

There is among other things a core extension that implements:
- a base controller: SelectedSiteController
- a template: Resources/Private/SelectedSite/DisplayNoSelectedPageWarning.html

The purpose of SelectedSiteController is to be inherited by other extensions which have a BE module needing a website (a page) to be selected in the page tree.

I’ve got two other extension which have a controller inheriting this base class SelectedSiteController

This class SelectedSiteController acts as a middleware before any action and get the currently selected website (page), or, if no page is currently selected, should display a basic template showing a ‘Please select a website’ message.

The SelectedSiteController class looks like this:

class SelectedSiteController extends ActionController
{
    protected function callActionMethod() {
       // looking for the current website
       // ...

       if(!$selectedWebsite) {
           $this->forward('displayNoSelectedPageWarning', 'SelectedSite', 'MyCoreExt');
       }

       parent::callActionMethod();
    }

    protected function displayNoSelectedPageWarningAction() {}
}

Then, I have this class in another extension :

class MyController extends SelectedSiteController {
    // various actions
}

If I don’t precise the extension in the ‘forward’ method, the child controller look for the template DisplayNoSelectedPageWarning.html in its own extension.

But if I define it explicitly, I’ve got this error:

(1/2) #1278450972 TYPO3\CMS\Extbase\Reflection\Exception\UnknownClassException

Class does not exist. Reflection failed.
in /var/www/typo3/public/typo3/sysext/extbase/Classes/Reflection/ReflectionService.php line 125
Author Olivier Massot
Created 12/08/21, 15:16
Updated 12/08/21, 15:16
Bug
New Orphaned File Abstraction Layer (FAL) TYPO3 CMS 10

#95330 'behaviour' => 'allowLanguageSynchronization' ignored on pages

We use “pages” with an own doktype as news pages, with an own model and repository (similar to the blog extension).

On translated pages, the TCA option ‘behaviour’ => ‘allowLanguageSynchronization’ is ignored, the sys_file_reference is always in default language 0, even if the translation behaviour is set to “custom value”.

The l10n_parent value of the translated values in sys_file_processedfile is always 0 – we tried to set the l10n_parent manually in database, then the translation of our FAL field works as expected (tried with custom fields and also with media field on plain TYPO3).

(this behaviour only happens on pages – tt_content is everything fine)

Author David Hoeckele
Created 09/23/21, 13:44
Updated 12/06/21, 14:11
Feature
Author Oliver Hader
Created 08/31/21, 13:46
Updated 11/29/21, 14:02
Task
Under Review Fresh Performance TYPO3 CMS 15

#111330 Check page access along the cached page ancestry

Most backend modules call BackendUtility::readPageAccess() with the show permission clause of the user, getPagePermsClause(Permission::PAGE_SHOW). As a clause is given, BackendUserAuthentication::isInWebMount() walks the page tree with it, one query per page level, instead of using the page ancestry resolved with one query since #111307. getRecordPath() then walks the page tree a second time for the record path, as the runtime cache of BEgetRootLine() differs in the prefix of the clause.

Proposal

  • isInWebMount() treats the show permission clause of the user like no clause and checks the page along its cached ancestry.
  • readPageAccess() takes the record path from the same ancestry, up to the first page not shown to the user.
  • Other clauses, listeners of the ModifyPagePermissionClauseEvent (#111328) and the deprecated getPagePermsClause hook keep the walk with the clause.

For a page on level seven, readPageAccess() then needs the record query and one ancestry query instead of two walks of the page tree.

Author Benni Mack
Created 11/03/26, 11:30
Updated 11/18/26, 19:16
Bug
Under Review Fresh TYPO3 CMS 14

#111489 Workspace preview of translated page fails with Unknown column '_TRANSLATION_SOURCE'

When previewing a non-default language page while a workspace is selected
(e.g. Staging + English), the frontend returns a 503:

An exception occurred while executing a query:
Unknown column ‘_TRANSLATION_SOURCE’ in ‘field list’

*How to reproduce*
1. Enable workspaces and create a Staging workspace.
2. Have a page with a connected-mode translation (e.g. German default,
English translation).
3. In the backend, select the Staging workspace.
4. Open the page module, select the translated language, and use
“View” / frontend preview.

Expected: translated page renders in workspace preview.
Actual: 503 with Unknown column ‘_TRANSLATION_SOURCE’.

*Root cause*
PageRepository::getPageOverlay() attaches the virtual property
_TRANSLATION_SOURCE (a Page object) to the page row.

versionOL() / getWorkspaceVersionOfRecord() then build the SQL SELECT
list from array_keys(purgeComputedProperties($row)).
_TRANSLATION_SOURCE was missing from COMPUTED_PROPERTY_NAMES, so it
leaked into the field list and MariaDB/MySQL rejected the query.

Note: SQLite may not always fail the same way; MariaDB/MySQL do.

*Suggested fix*
Add ‘_TRANSLATION_SOURCE’ to PageRepository::COMPUTED_PROPERTY_NAMES
so it is stripped before workspace overlay queries, same as
_LOCALIZED_UID / _ORIG_uid / etc.

This is unrelated to #108496 / https://review.typo3.org/c/Packages/TYPO3.CMS/+/92672
(which only handles _TRANSLATION_SOURCE in MenuProcessor JSON encoding).

Author Himanshu Ramavat
Created 11/18/26, 12:03
Updated 11/18/26, 12:05
Bug
Accepted Fresh Backend User Interface TYPO3 CMS 14

#111333 selectMultipleSideBySide available/selected lists are blank on Safari iOS

## Summary
FormEngine `selectMultipleSideBySide` shows empty white list boxes on Safari on iPhone / iOS. Labels, filter input and move buttons render, but `<option>` items are not visible.</option>

## Environment
- TYPO3: 14.3.7
- PHP: 8.5
- Browser: Safari 27 on iOS / iOS 27 / iPhone

Author Vivek Solanki
Created 11/03/26, 14:57
Updated 11/18/26, 07:01
Epic
Accepted Fresh Authentication

#110448 Modernizing TYPO3 Authentication — Phase 1: The Authentication Pipeline

Why we are doing this

Authentication is some of the oldest code we have in TYPO3. AbstractUserAuthentication dates back to the early 2000s, and it shows: it is a god class that fetches sessions, sniffs login credentials out of POST data on every single request, runs a string-typed ““auth service”” chain (T3_SERVICES, numeric priorities, addService()) that predates dependency injection by 15 years, sets cookies, writes log entries and mutates global state ($GLOBALS['BE_USER'], $GLOBALS['TSFE']->fe_user) along the way.

The practical pains coming out of this:

  • Login is buried in start(), not explicit. In the backend, any request carrying the right POST fields triggers authentication — there is no login route, which makes SSO integrations, rate limiting and auditing unnecessarily hard. In the frontend, ““the login form can live on any page”” is a feature everybody uses (felogin posts back to the page you are on) and we keep it — but the way it works today is the same credential sniffing on every single request, impossible to reason about or extend.
  • Extending authentication means registering a legacy service with magic numbers (what does processLoginData returning 200 mean again?), no types, no DI. Every SSO/LDAP extension author suffers through this.
  • MFA is bolted on via an exception (MfaRequiredException) that only the backend knows how to catch. Frontend MFA is structurally impossible right now.
  • No concept for machine/API authentication. Tokens, service accounts, headless — everything has to fake being a cookie-based user.
  • You cannot test it. The 235-line checkAuthentication() method had no dedicated test coverage until we wrote characterization tests in Phase 0.

What Phase 1 introduces

Phase 1 puts a small, boring, typed pipeline underneath the existing behavior. No behavior changes for users, no changes for extensions — the new code is internal@ for now and the old APIs keep working.

The new pieces and what each one is responsible for:

  • @Principal@ — an immutable value object answering ““who is authenticated?”“: user id, realm, username, groups, claims, how they authenticated (password? session?) and when. This is the thing the rest of the system should care about — not a 1500-line user object.
  • @Realm@ — answers ““authenticated where?”“: backend, or frontend per site. It also knows which routes don’t need authentication (currently the hard-coded public route list from the middleware).
  • @AuthenticatorInterface@ — one strategy for proving identity. Each authenticator says whether it applies to a request (supports()) and returns exactly one of four typed results: Authenticated (here is a Principal), Challenge (I need more — e.g. an MFA response to redirect to), Failure (wrong password) or NotApplicable. This replaces the numeric service return codes.
  • @AuthenticationManager@ — runs the DI-tagged authenticators in priority order, first conclusive result wins. That’s the whole class. Registering a custom authenticator becomes: implement an interface, tag a service.
  • Native SessionAuthenticator + PasswordAuthenticator (BE and FE) — reimplement what the legacy AuthenticationService chain does for a plain install: session lookup, password check with rehash-on-login, rate limiting, request-token check, session fixation protection. In the backend, password handling is now bound to the login routes. In the frontend, the trigger stays what it is today — an explicit logintype=login parameter on whatever page the login plugin sits on, because login-on-any-page is a feature, not an accident. The difference: the trigger is now an explicit, testable condition (parameter + signed request token) instead of code buried deep in the user object.
  • @LocalIdentityProvider@ — the one place that looks up be_users/fe_users rows (enable fields, storage pids). Authenticators don’t build queries anymore.
  • Legacy wrappers + detector — the entire old service chain is wrapped as one authenticator. As soon as an extension registers an auth service, we detect that and run the full legacy chain exactly as before. Clean installs run the native path. Nothing breaks either way.
  • MFA as a Challenge — instead of throwing an exception, the pipeline returns a redirect/401 response. Same backend behavior as today, but the mechanism is realm-agnostic, which is what will eventually enable frontend MFA.
  • @UserAspect@ — becomes an immutable snapshot and can be built from a Principal directly, cutting the core→frontend dependency it had.

The honest downsides

  • We temporarily have two implementations of login (native + wrapped legacy chain). That’s the price for not breaking a single existing SSO extension. Characterization and functional tests pin both paths to the same behavior.
  • The native path has two deliberate behavior changes: credentials POSTed to arbitrary backend URLs are ignored (BE login belongs to the login routes; frontend login-on-any-page is untouched), and MFA redirects instead of throwing. We consider both features, not regressions.
  • The pipeline is internal@ in 15.0. Extension authors should look, comment — but not build on it yet. It becomes public API once the login endpoints (end of Phase 1) settle.
  • The god classes are still there. Phase 1 changes who is in charge, not yet who holds the data.

Where this is going

Phase 1 is the foundation everything else plugs into:

  • Phase 2 — Identity & machine auth: sys_identity account linking, sys_token + TokenAuthenticator with scopes, service accounts (bye _cli_), a stateless api realm for headless.
  • Phase 3 — Federation: an OIDC sysext (login button, PKCE, back-channel logout), passkeys/WebAuthn, frontend MFA — all just more authenticators and challenges on the same pipeline.
  • Phase 4 — Decoupling: $GLOBALS['BE_USER'] replaced by Principal + Context, permissions extracted from the user class, preview/simulation as derived Principals instead of in-place mutation.
  • v16: the legacy service chain, checkAuthentication() and friends are removed. Everything new lands and gets deprecated within v15 — v16 only deletes.
Author Benni Mack
Assigned to Benni Mack
Created 08/12/26, 11:11
Updated 11/17/26, 16:10
Feature
New Fresh Site Handling, Site Sets & Routing

#111468 Provide a way to shim site-set dependency resolution

It is mentally very hard to work with site-sets. Switching from explicit to implicit ordering was a tough choice by TYPO3. I developed more than 100 TYPO3 websites and get stuck again and again, so maybe a little be too tough and it might be possible, that TYPO3 made a bad decision (or i am just way too stupid) and i would like to propose a way to mitigate the issues.

If an extension author would be able to “shim” the dependencies of third-party site-sets, it would be very easy to solve various dependency ordering issues. Just like it was added to various module loaders in the past, when we had the same problems with third-party-modules and the combination of those.

That could be done by providing a PSR-14 Event. I will try to replace the service class for now (to add the shimming for my use-case) but a generic way to improve interoperability between third-party site-sets and my own site-sets would be the preferred way.

Why a PSR-14 Event? Because it can be added without introducing any breaking change, so this could be backported to V14 and V13 as well. As alternative, Set Overrides could work but that would be a breaking change for a Major Release.

More information might be found here https://talk.typo3.org/t/site-sets-are-killing-me-dependency-ordering-hell/6738/2, where i initially reached out for help

Author Philipp Wrann
Created 11/16/26, 16:34
Updated 11/17/26, 13:22
Task
Under Review Fresh Extension Manager TYPO3 CMS 15

#111471 Move excludeForPackaging to Extension Manager configuration

The setting $GLOBALS[‘TYPO3_CONF_VARS’][‘EXT’][‘excludeForPackaging’] is only used by the Extension Manager, which excludes matching files and folders when an extension is downloaded as ZIP archive.

It should move to the extension configuration of EXT:extensionmanager (ext_conf_template.txt), so it lives next to the other Extension Manager settings. This also removes the last setting of the $GLOBALS[‘TYPO3_CONF_VARS’][‘EXT’] section.

A silent configuration upgrade migrates a custom value from config/system/settings.php to $GLOBALS[‘TYPO3_CONF_VARS’][‘EXTENSIONS’][‘extensionmanager’][‘excludeForPackaging’].

Author Benni Mack
Created 11/17/26, 11:53
Updated 11/17/26, 11:54
Bug
Under Review Fresh Backend API TYPO3 CMS 14

#111268 SuggestWizardDefaultReceiver::queryTable() exhausts memory when maxItemsInResultList is high

Problem

Searching in the suggest wizard of a group field runs into PHP’s memory limit when
@maxItemsInResultList@ is set well above 50 and the search term matches 50 or more
records. Terms with fewer than 50 matches work.

Steps to reproduce

Verified on a fresh cms-base-distribution v14.3.10 (SQLite, no extensions):

  1. Create 122 content elements on page 1 whose header contains “ter” (e.g. “Water 1” to
    “Water 122”) and 22 whose header contains “obe” (e.g. “Globe 1” to “Globe 22”).
  2. Set page TSconfig on page 1:
TCEFORM.tt_content.records.suggest.default.maxItemsInResultList = 9999
  1. Create a content element of type “Insert records” on page 1 and search “ter” in the
    suggest field of “Items”.

Expected

The suggest list shows the matching records.

Actual

  • Searching “obe” (22 matches) returns all 22 records.
  • Searching “ter” (122 matches) fails after a few seconds with
    “Allowed memory size of 134217728 bytes exhausted” (memory_limit = 128M).
  • With memory_limit = 512M, SQLite fails instead with
    “General error: 1 Expression tree is too large (maximum depth 1000)”, which shows
    the WHERE clause growing on every call.
  • With maxItemsInResultList = 50, “ter” returns 50 records and uses 26 MB.

Cause

queryTable() fetches 50 rows per call and calls itself for the next page
(SuggestWizardDefaultReceiver.php#L185):

if (count($rows) < $this->maxItems && $allRowsCount >= $maxQueryResults && $recursionCounter < $this->maxItems) {

Three things combine:

  1. The recursion does not stop when the results run out. $allRowsCount is the total
    number of matches and does not change between calls, and $rows only holds the
    current page. Once the matches are used up, every further call returns no rows and
    still calls itself again, until $recursionCounter reaches maxItems.
  2. The call limit is compared with the item limit. $recursionCounter < $this->maxItems
    allows up to maxItemsInResultList calls (9999 in the example) instead of
    ceil($allRowsCount / 50) pages.
  3. The WHERE clause grows on every call. queryTable() calls prepareSelectStatement()
    on every call
    (L139 which adds the search constraints and searchCondition to $this->queryBuilder
    again, with new parameters each time. Each call then clones a larger query builder,
    so memory use grows with the square of the call depth.

With 122 matches and maxItemsInResultList = 9999, three calls would be enough; the
code keeps calling itself until memory runs out.

A second effect of the recursion: results are merged with array_merge($tmp, $rows),
which puts later pages in front of earlier ones. When the controller then trims the
list to maxItemsInResultList, it removes the records that sort first. With 122
matches and maxItemsInResultList = 60, the list does not start with the first
record in sort order.

Custom receivers that extend the class and call parent::prepareSelectStatement() are
affected the same way.

Possible fix

  • Call prepareSelectStatement() once, before fetching.
  • Replace the recursion with a loop over pages that stops when a page returns fewer than
    50 rows, or when maxItems records have been collected.
  • Keep the sort order of the result.
  • Drop the separate count query, as #106981 proposed.

Note: stopping when a page yields no accepted rows (the approach in the abandoned
change https://review.typo3.org/c/Packages/TYPO3.CMS/+/89912) would end the search early
if all 50 rows of a page fail checkRecordAccess(). The stop condition should depend on
the rows fetched, not on the rows accepted.

A patch with functional tests along these lines will follow on Gerrit.

Workaround

Keep maxItemsInResultList at 50 or lower. That limits the number of calls to roughly
@maxItemsInResultList / 50@.

Related

  • #106981 (rejected) and its abandoned change 89912 touched the same recursion but
    targeted query count and result limiting, not the memory problem.
Author Johannes Schlier
Created 10/28/26, 14:15
Updated 11/16/26, 16:36
Task
Under Review Fresh Backend User Interface TYPO3 CMS 15

#111466 Use flag icons optimized for small sizes

The flag icons are rendered from the detailed “flag-icons” set. At the 16px size used in the backend most flags lose their details and look blurry, especially flags with coats of arms.

Switch to “flagpack” (MIT), which is drawn for 16×12px, and render the flags with rounded corners and a subtle outline. The overrides and the TYPO3-specific flags (colours, rainbow, multiple, en-us-gb, …) are redrawn in the same simplified style.

  • Flags are still shipped as WebP (now 64px instead of 32px), so they stay sharp on high resolution screens.
  • All existing flags-* identifiers and file names are kept, site configurations and templates continue to work.
  • New flags: gb-ukm, bq-bo, bq-sa, bq-se.
  • Raster images keep pages listing hundreds of flags (e.g. the site language configuration) as fast as before. SVG was evaluated, but rasterizing thousands of SVG images caused a noticeable delay.
Author Alexander Hahn
Created 11/16/26, 14:06
Updated 11/16/26, 14:29
Bug
Under Review Fresh Backend User Interface TYPO3 CMS 13

#107455 File/FAL fields ignore onChange='reload' configuration in FormEngine

Current Behavior

When configuring a file/FAL field (type=‘file’) with onChange=‘reload’ in TCA, the form does not reload when files are added, removed, or reordered. This differs from the behavior of other field types (input, select, etc.) where onChange=‘reload’ triggers a form reload with a confirmation dialog.

Expected Behavior

File/FAL fields should respect the onChange=‘reload’ configuration and trigger a form reload (with confirmation if configured) when the field value changes, consistent with other field types.

Steps to Reproduce

1. Add onChange=‘reload’ to a file field in TCA:

$GLOBALS['TCA']['pages']['columns']['media']['config']['onChange'] = 'reload';

2. Edit a page record in the backend
3. Add or remove a file in the media field
4. Observe that no reload confirmation appears and the form doesn’t reload

Author Wolfgang Klinger
Created 09/14/25, 13:37
Updated 11/16/26, 13:39
Bug
Under Review Fresh TYPO3 CMS 14

#107447 There are no landmarks/regions present beside the navigation (wasWCAG: Language change functionality)

Description

It is hard for non visual users to effectively navigate to a certain section within the viewport aside from the toolbar or the iframe since there are no landmarks present.
This affects e.g. the language setting area or the widgets on the dashboard.
Example: when navigating within the iframe, navigation starts with “Layout”, Language Setting area is lost.

Reproduction steps

open Pages
navigate via Screenreader through iframe

Impact

Target group

visual

Severity

high

Frequency

very common

Definition of Done

WHEN I use a desktop screenreader (NVDA, JAWS, VoiceOver) AND
I use the tab key to enter the web browser window
I HEAR all main functional areas (main, header, etc,) are discoverable as landmarks or regions

Recommended solution

Either assign aria-regions to the big areas of focus, within the iframe and aside from it or alternatively use semantic html elements. Keep in mind role=region needs label to be identifiable.

Examples:

header (could be semantic element)
pagetree (aria-region)
language setting (aria-region)
Workspace/main (could be semantic element)

References / WCAG 2.2 Reference

2.4.1. Bypass Blocks

Prioritizing the removal of the issues found

must have

Screencast or Screenshot

—-

original message:

The user is on the “Page” module point. The vertical menu of the page hierarchies and the work area open to the right of it.

The language comparison functionality now appears above, displayed in a drop-down menu that is filled with the Layout selection by default.

Layout is specified proportionally as a frame, but “Language comparison” cannot be accessed via frames. There is no consistency here.

Via the navigation, the page effectively starts with “Layout” when the field is selected. It is unclear where the user is located and what the menu/function belongs to. Add heading desired. (H1). Layout and Language Comparison are somewhat lost.

Solution idea:

Add a structuring element to assign the functionality to the work area. Include the language change in a tab?

The Layout and Language Comparison fields must be output with the same attributes (frame?).

Expected result:

The user navigates from the page tree to the workspace and is given the option of displaying the page in English or French. The menu is located where expected and is integrated into the workspace.

internal reference: ITZBUNDPHP-4985

Author Christian Rath-Ulrich
Created 09/12/25, 08:40
Updated 11/16/26, 12:41
Bug
Under Review Fresh FormEngine aka TCEforms TYPO3 CMS 13

#109011 Content edit permission error on page settings edit

Environment:

- user in user group with `view page`, `edit page` and `add page` permissions but no `edit content` or `delete page`

If user wants to open the page settings of a page with an already existing related media file in resources tab, an error message with id 1437679657 is thrown.
This happens, because the check at https://github.com/TYPO3/typo3/blob/14.0/typo3/sysext/backend/Classes/Form/FormDataProvider/DatabaseUserPermissionCheck.php#L114-L134 will called twice. Once with `tableName = 'pages'` which is successful and second failing run with `tableName = sys_file_reference` because of the related media file. If I delete the related media file as admin and open the page settings again there is no second run and user can open without an error message.

This error happens since TYPO3 v6 but only now have we had time to take a closer look at it :) Tested in latest v12 and v13.

Author Roman Teitge
Created 02/18/26, 19:53
Updated 11/16/26, 10:16
Feature
Under Review Fresh WebHooks - Incoming = Reactions + Outgoing

#111024 Allow webhook messages to target specific webhooks

EXT:webhooks delivers every webhook message to all sys_webhook records configured for the message’s trigger. The code dispatching a message has no way to restrict delivery to a subset of these records.

This is fine for system-wide triggers like “login error” or “file added”, where every configured endpoint is interested in every occurrence. It is not sufficient for triggers whose payload belongs to a specific configuration in the backend – for example a form submission, where an editor wants the data of one particular form to reach one particular endpoint, and personal data must not be sent to endpoints that are not supposed to receive it.

Introduce TYPO3\CMS\Core\Messaging\TargetedWebhookMessageInterface, extending WebhookMessageInterface, with one method:

public function getTargetWebhookIdentifiers(): array;

returning the identifiers (UUIDs) of the sys_webhook records the message is meant for.

WebhookMessageHandler checks for the interface. If a message implements it, only records that
  • are configured for the message’s trigger,
  • are active (default restrictions), and
  • have an identifier listed by the message
    receive the request.

Following rules may apply:

  • An empty list delivers nothing (fail-closed).
  • Identifiers of records configured for another trigger are ignored, so the payload of one trigger can never reach an endpoint configured for another.
  • Listed identifiers without a matching active record are skipped and logged as a warning.
  • Messages that do not implement the interface are delivered exactly as before.

The target identifiers are properties of the message, not part of the JSON body, and survive serialization when the message is routed to an asynchronous transport.

Author Josua Vogel
Created 10/12/26, 06:12
Updated 11/15/26, 16:16

Showing results 526 to 550 of 1561