Issue Bug Closed

#102015Incorrect behavior of TCEFORM treeConfig.startingPoints

Robert von Hackwitz 09/22/23, 12:08 updated 11/13/26, 08:53

Description

When you create a category tree by assigning the category mount to a group of non-privileged users
and at the same time using the TCEFORM treeConfig.startingPoints option, only the parent category is displayed.
In the example (of which I attach screenshots) I created a tree of categories and subcategories,
I assigned a category in the category mount point of the BE user group and in the root page of the site I
used TCEFORM.sys_category.parent.config.treeConfig.startingPoints=2581 option (where 2581 is the Uid of the “Main Root category” category).
When the non-privileged user belonging to the group tries to create a new category in the “parent” tree only “Main Root Category” appears.
If I don’t use TCEFORM.sys_category.parent.config.treeConfig.startingPoints=2581 or if the user is a TYPO3 admin then TCEFORM.sys_category.parent.config.treeConfig.startingPoints=2581 works fine

Reviews 1

Files 3

History 1425

  1. Robert von Hackwitz commented <span data-processor="relativetime" data-value="2023-10-02T08:31:24+00:00" > 10/02/23, 08:31 </span>

    Hi,
    after a bit of investigation I think there is something wrong in
    TYPO3\CMS\Backend\Security\CategoryPermissionsAspect around lines 70-86

                   foreach ($startingPoints as $startingPoint) {
    
                            if (!in_array($startingPoint, $categoryMountPoints)) {
                                $shallRepopulateTree = true;
                                break;
                            }
                            $uidsInRootline = $this->findUidsInRootline($startingPoint);
                            if (empty(array_intersect($categoryMountPoints, $uidsInRootline))) {
                                $shallRepopulateTree = true;
                                break;
                            }
                    }

    With the for loop we check whether startingpoints are present in the user’s categorymounts,
    but the starting points can be (as in my case) subcategories of a category mount points so the if statement is false, the variable $shallRepopulateTree is set to true and the cycle is interrupted, causing the reconstruction of the tree without the categories corresponding to the starting point

  2. Uwe Wiebach commented <span data-processor="relativetime" data-value="2024-06-06T09:50:22+00:00" > 06/06/24, 09:50 </span>

    It’s definitely this part, because if I comment it out, the tree is built up correctly.

    So I dug deeper:
    If the starting point (or back then the rootUid) is a category mount point before the changes for https://forge.typo3.org/issues/95037 the tree wasn’t touched.

                        if (in_array($dataProvider->getRootUid(), $categoryMountPoints)) {
                            return;
                        }

    Now its rootline is also checked and that’s where it all goes wrong…

    I’m sure this needs more testing but in my case this worked:

                    foreach ($startingPoints as $startingPoint) {
                        if (in_array($startingPoint, $categoryMountPoints)) {
                            $shallRepopulateTree = false;
                            break;
                        }
                        $uidsInRootline = $this->findUidsInRootline($startingPoint);
                        if (!empty(array_intersect($categoryMountPoints, $uidsInRootline))) {
                            $shallRepopulateTree = false;
                            break;
                        }
                        $shallRepopulateTree = true;
                    }

    Maybe this $shallRepopulateTree = false; isn’t needed as false is the inital value.

    EDIT:
    This only works if a starting point is set… So I changed it to:

                    foreach ($startingPoints as $startingPoint) {
                        if ($startingPoint) {
                            if (in_array($startingPoint, $categoryMountPoints)) {
                                $shallRepopulateTree = false;
                                break;
                            }
                            $uidsInRootline = $this->findUidsInRootline($startingPoint);
                            if (!empty(array_intersect($categoryMountPoints, $uidsInRootline))) {
                                $shallRepopulateTree = false;
                                break;
                            }
                            $shallRepopulateTree = true;
                        }
                    }
  3. Uwe Wiebach commented and changed attached patch_102015.diff <span data-processor="relativetime" data-value="2024-08-30T11:36:56+00:00" > 08/30/24, 11:36 </span>

    Added my changes as a patch…

  4. Uwe Wiebach commented and changed attached patch_102015.patch <span data-processor="relativetime" data-value="2024-10-28T14:26:36+00:00" > 10/28/24, 14:26 </span>

    Unfortunately, I have only just realized that the patch file was faulty. This one should be correct…

  5. Andreas Kienast commented and changed status New Accepted <span data-processor="relativetime" data-value="2025-01-16T13:15:12+00:00" > 01/16/25, 13:15 </span>

    Unfortunately, the attached patch is not sufficient. It is possible to configure multiple starting points, more than a backend user may have access to. The loop aborts after the first successful run, ignoring additional starting points, which will allow to “break out” of configured mounts.

  6. Gerrit Code Review pushed patch set 5 for main to change 87780 status Accepted Under Review <span data-processor="relativetime" data-value="2025-01-16T15:35:29+00:00" > 01/16/25, 15:35 </span>

  7. Florian Rival commented <span data-processor="relativetime" data-value="2026-05-11T12:48:43+00:00" > 05/11/26, 12:48 </span>

    The bug is still present, the current patch is abandoned, are there any plan to correct it ?

  8. Benni Mack commented and changed status New Closed <span data-processor="relativetime" data-value="2026-11-13T08:53:27+00:00" > 11/13/26, 08:53 </span>

    Fixed with #106559