Review main Merged

93721[BUGFIX] Do not set stability for known framework packages

Helmut Hummel CR +2 V +2 Easy +3 −3 updated 05/08/26, 09:36

Commit message

Although we should not support third party extension
announcing themselves as framework packages, we allowed
this in the past without errors.

This is left as is, but the version of the extension
is now only set to `Typo3Version`, when the extension
is known as Framework extension, regardless of the
announced type.

Additionally setting stability is now only changed
for non framework packages, since the stability
of framework packages is already incorporated
in the version number.

Review

  • Code review CR +2 Done
    • Helmut Hummel +2
    • Oliver Klee +1
    • Garvin Hicking +1
    • Christian Kuhn +1
  • Verified V +2 Done
    • Helmut Hummel +2
    • Oliver Klee +1
    • core-ci +1 CI

Issues 1

Files 1

Review log 2529

Show 24 earlier messages
  1. Helmut Hummel uploaded patch set 1 04/17/26, 17:14

  2. Helmut Hummel updated the commit message in patch set 2 04/17/26, 17:17

  3. core-ci voted on patch set 1 V -1 04/17/26, 17:24 Core CI is not happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/100403

  4. core-ci voted on patch set 2 V -1 04/17/26, 17:29 Core CI is not happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/100405

  1. Helmut Hummel uploaded patch set 3 V -1 04/17/26, 17:54

  2. core-ci voted on patch set 3 V +1 04/17/26, 18:05 Core CI is happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/100406

  3. Oliver Klee voted on patch set 3 V +1 04/19/26, 11:17

  4. Oliver Klee voted on patch set 3 CR +1 04/19/26, 11:18

  5. André Buchmann commented on patch set 3 CR -1 04/19/26, 19:57

    Package.php:160

    I think we should throw an exception here instead of trying to fix an invalid extension on the fly. However, if I replace the todo with a InvalidPackageMainifestException, the exception is captured silently, and the extension unregistered. No log entry in the files/backend module, no hint in the extension manager or in the Scanners of the Install-Tool/Upgrades. So the integrator has no clue, why is extension wouldn't show up in all those places. Tested it with this: throw new InvalidPackageManifestException('Invalid type "%s" used in composer manifest in package "%s". Please use "typo3-cms-extension" instead.', 1776627733); On the other hand, if we keep the "fix" it way, the integrator also doesn't get notified in any way of the invalid definition.
  6. Helmut Hummel commented on patch set 3 04/20/26, 05:01

    Package.php:160 Reply

    All good points. However: do you think integrators set the extension type to framework? Is that something useful? What do they want to achieve? What breaks if we set this back to extension?
  7. Helmut Hummel commented on patch set 3 04/20/26, 05:06

    Package.php:160 Reply

    If changing the package type is disputed, I can easily leave this out, as the more important fix is below. I'm still interested in learning why one sets the extension type of third party extensions to framework...
  8. André Buchmann commented on patch set 3 04/20/26, 06:35

    Package.php:160 Reply

    Maybe integrator is the wrong wording here, replace it with (inexperienced?) extension developer, maybe even AI ;-) I prefer throwing an exception over fixing it, but open to change my mind if others are fine with the fixing. Is it possible to write a log entry in the PackageManager::registerPackagesFromConfiguration() try-catch to have at least some notification about the error?
  9. Helmut Hummel commented on patch set 3 04/20/26, 07:29

    Package.php:160 Reply

    Again: What do you think or assume breaks for the "inexperienced" developer who created the code with AI, when we fix the wrong package type here? As far as I know, there is nothing you can't achieve with a regular extension. What would be breaking though, if we'd throw an exception here, as previously "AI generated code" would stop working.
  10. Helmut Hummel commented on patch set 3 04/20/26, 07:34

    Package.php:177

    @andy.schliesser@gmail.com It is however a good catch, that throwing InvalidPackageManifestException will silently unregister the extension. This however is true for *any* usage of InvalidPackageManifestException. If this is considered to be improvable (like logging reasons for the error), this can (and for my part should) be done in a different change
  11. Benjamin Franzke commented on patch set 3 04/20/26, 09:44

    I know, this is getting complex again… I found another case which leads to the same exception, which I've found that at least two extensions (EXT:sessionplanner and EXT:sf_event_mgt) used in the part, see https://forge.typo3.org/issues/109626#note-7 I don't know if that is wrong, but it would help to avoid suffixing the stability flag as shown in https://review.typo3.org/c/Packages/TYPO3.CMS/+/93745 to keep previous behavior.
  12. Helmut Hummel commented on patch set 3 04/20/26, 11:01

    Reply

    did not find usage of that in sf_event_mgt any hints? Other than that. IIRC using stability flags in ext_emconf.php might have worked, but were never supported/ documented, were they?
  13. Benjamin Franzke commented on patch set 3 04/20/26, 11:15

    Reply

    > did not find usage of that in sf_event_mgt. any hints? Has been in dev branches (it was called "typo3-v13" back then) and I had that lying around, the sha1 is: https://github.com/derhansen/sf_event_mgt/blob/34d8e1dc1c18e66e98cc68802e17863d23bd092a/ext_emconf.php#L19-L23 > Other than that. IIRC using stability flags in ext_emconf.php might have worked, but were never supported/ documented, were they? I don't know if they were documented, but I'm also not aware we documented any restrictions. So people probably could write anything in there. I guess ideally we'd "fallback" to whatever is provided as version string as-is, if a parsing exception happens. Hope this helps.
  14. Benjamin Franzke commented on patch set 3 04/20/26, 11:24

    Reply

    btw, `sf_event_mgt` does the same in the "typo3-v14" branch: https://github.com/derhansen/sf_event_mgt/blob/73af68a9bfd1d329d97c9b892589da0679820e40/ext_emconf.php#L15-L16 but I think it doesn't fail there because it defines the same way of suffixing the version number in `composer.json` which probably overwrites the ext_emconf value (didn't check that thoroughly though): https://github.com/derhansen/sf_event_mgt/blob/73af68a9bfd1d329d97c9b892589da0679820e40/composer.json#L5
  15. Helmut Hummel commented on patch set 3 04/20/26, 16:49

    Reply

    > I don't know if that is wrong, but it would help to avoid suffixing the stability flag as shown in https://review.typo3.org/c/Packages/TYPO3.CMS/+/93745 to keep previous behavior. As a quick fix, this is totally fine with me. Let's get this in then first and discuss this later
  16. André Buchmann commented on patch set 3 CR 0 04/20/26, 20:28

    Package.php:160 Reply

    I don't know the reason why anyone would define an extension as typo3-cms-framework. > What breaks if we set this back to extension? Nothing. > Although we should not support third party extension announcing themselves as framework packages, we allowed this in the past without errors. Throwing an InvalidPackageManifestException would unregister the extension now and therefore this would be a breaking change and shouldn't happen anymore for 14.3? If so, can we add a (deprecation) log entry that the type has been fixed for extension xyz. Kinda like the automatic TCA migrations?

    Package.php:177 Reply

    I agree that this should be a different patch to apply logging in the PackageManager. I've created a new issue: https://forge.typo3.org/issues/109659
  17. Helmut Hummel uploaded patch set 4 CR +1 V +1 04/27/26, 11:33

  18. Helmut Hummel commented on patch set 4 04/27/26, 11:36

    Package.php:160 Reply

    I realised that this does not matter for this change. The important part is, to only set the version (and state) for packages we know about. The rest might be left as is or changed in a different effort.
  19. core-ci voted on patch set 4 V +1 04/27/26, 11:53 Core CI is happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/101167

  20. Garvin Hicking voted on patch set 4 CR +1 04/27/26, 12:37

  21. Helmut Hummel commented on patch set 4 04/28/26, 10:03

    Reply

    Done
  22. Oliver Klee voted on patch set 4 CR +1 V +1 04/28/26, 10:51

  23. Christian Kuhn voted on patch set 4 CR +1 05/02/26, 11:14

  24. Helmut Hummel voted on patch set 4 CR +2 V +2 05/08/26, 09:36

  25. Helmut Hummel merged patch set 5 05/08/26, 09:36