[meta] Address PHPUnit 9 deprecations in Drupal 10 for forward-compatibility with PHPUnit 10

Re-parenting to the correct meta (we can do this in D10 minors).

Symfony have started the discussion on how to adjust the PHPUnit-Bridge package for PHPUnit 10, https://github.com/symfony/symfony/issues/49069.

prophecy-phpunit too, https://github.com/phpspec/prophecy-phpunit/pull/41

Updated IS.

Gábor Hojtsy → credited larowlan → .

PHPUnit 10 will be released on February 3, 2023, in two days as per https://thephpcc.social/@sebastian/109788385318681908

Updated issue summary with this info.

Hat tip to @larowlan for the info.

Link to changelog

We need to wait for Symfony and Prophecy to catch up first, we can't really move forward on this until they are compatible too.

Concurrently with PHPUnit 10, PHPUnit 9.6 will be released, too. That release will deprecate quite a number of things that we need to work on before we can tackle PHPUnit 10, too: https://www.drupal.org/project/issues/search?issue_tags=PHPUnit%209.6 →

PHPUnit 10 is out, https://phpunit.de/announcements/phpunit-10.html

An annoucement worth reading.

Interesting indeed, thanks for linking that. Given that PHPUnit 9's end of life has yet to be determined (but the blog post implies it will be longer than the end of bugfix support in 2024) and that we plan to release Drupal 11 (also perhaps in 2024), we can consider staying on PHPUnit 9 in the Drupal 10 cycle while clearing the way to upgrade to PHPUnit 10 for Drupal 11.

In the meantime, we can still upgrade to PHPUnit 9.6: 📌 Update to PHPUnit 9.6.7 Fixed

Fair bit going on in that post, thanks for sharing.
.
We're going to need to wait for prophecy and the symfony bridge to have support too

Targeting D11 sounds like a reasonable plan

If PHPUnit release schedule goes back to yearly as it was in the past, https://github.com/sebastianbergmann/phpunit/milestones, PHPUnit will be on its 11 release by the time D11 ships. And PHPUnit 10 will likely be EOL in Feb 2025, while D11 will be still on. Combined with PHP evolution itself, PHPUnit 10 may become obsolete during D11 lifecycle. Fun.

I feel like the issue title is no longer representing what's being discussed here, I suggest to revert to an updated version of the prior one.

Added reference to https://github.com/sebastianbergmann/phpunit/issues/5201 in the IS

rebase

phpspec/prophecy-phpunit 2.1.0, together with phpspec/prophecy 1.18.0, are meant to be PHPUnit 10 compatible

The MR now runs all the test in PHPUnit 10, so we can track impact when the related items fix things.

Some fixes after 🐛 '@requires externalCommand' is not parsed in PHPUnit 10 Fixed .

Prophecy have PHPUnit 10 compatibility fixes in, but have not released yet.

Symfony PHPUnit-Bridge is rather silent, I have opened a PR https://github.com/symfony/symfony/pull/50371, its diff is used to get the MR here going. But that is mostly experimental.

Fixing attribution.

Don't see any discussion on it but our test suite registration is no longer supported in PHPUnit 10.

https://github.com/sebastianbergmann/phpunit/commit/36354a9cdc69ecabd20a...

Might be a really cool solution in the current patch I'm missing though.

#95 since there's absolutely nothing in the current patch that thoughtfully addresses test suites, I started looking for miracles... :)

But ended up, in my understanding, finding that we are not 'using' test suites at all even now. AFAICS. run-tests.sh builds list of individual tests to be run in parallel, but does not use test suites (in PHPUnit's sense) for that, rather our own-baked definition of test groups via matching test classes namespaces. It does so in TestDiscovery::getPhpunitTestSuite().

Since core is only using run-tests.sh, I wonder what are we defining TestSuites for? Just for those that want to use the suite filtering when running PHPUnit via its own CLI? If that's the case then I think we should drop this - at least start deprecating them now.

📌 Deprecate Test Suites, no longer available in PHPUnit 10 Needs review shows IMHO that test suites are just cruft currently.

Re: #98 failures.

It looks like 📌 Split PHP image into php(cli/apache) and yarn(node/nightwatch) Needs review removed the patch command from images.

Spent a bunch of time digging into the discovery process related to the test suites and that sent me down a lot of rabbit holes. The test suite part I did not enjoy but it ended with me having a mostly working unit test suite so that's great! Of course mondrake was ahead of me on basically all of those changes.

Tried to capture all of my findings and the various work happening in an updated IS.

Will capture some issues for some of the things I found that I think can be addressed in the short term.

Great writeup, thanks @neclimdul!

Hide patch file, try to get this into the MR

Found another interesting issue exposed by the PHPUnit 10 updates.

🐛 DependencySerializationTrait depends on removed __PHPUNIT_BOOTSTRAP global Active

The correct fix probably needs some architecture discussion because it would require changes to DependencySerializationTrait which affects almost everything.

phpspec/prophecy-phpunit 2.1.0 is now in

Opened 📌 Upgrade PHPUnit to 10, drop Symfony PHPUnit-bridge dependency Fixed to discuss removal of Symfony's PHPUnit bridge in case it is not getting compatible with PHPUnit 10 in time for Drupal 11 release.

Amazing work here :) It would be great to clean up the issue summary to current priorities and remaining work. I am not sure I can tell in each point whether its done or not. For example the Symfony PHPUnit bridge question is handled in 📌 Upgrade PHPUnit to 10, drop Symfony PHPUnit-bridge dependency Fixed . This seems to be the biggest backend dependency update for Drupal 11 apart of Symfony 7 and it would be great to have this in place sooner than later so that contrib modules can update to and test with the new version.

Also as per the comments above PHPUnit 11 is about to be released later this week on Friday. As per https://phpunit.de/supported-versions.html that also means end of bugfix support for PHPUnit 9 that we are using now. While PHPUnit 9 and 10 EOLs have not been determined yet (again as per that page), people argued above that bugfix support for PHPUnit 10 will also end in 2025 February, early into the Drupal 11 release cycle. So if at all possible, we should get to PHPUnit 11 in time for Drupal 11 :) (not in this issue, but this is a prerequisite).

Is there any way the committer team can better help? Is framework manager review needed on certain areas? Should the issue be broken out to (further) smaller changes?

Updated the withConsecutive() options.

Opened 🌱 [meta] Support PHPUnit 11 in Drupal 10 Postponed as postponed and put it in the issue summary as a followup :) That allows us to track it on 🌱 [meta] Add compatibility for the latest major and minor versions of dependencies to Drupal 10 Active too.

Updated the IS with the completed tasks.

Let’s be honest.

@mondrake Per https://www.drupal.org/blog/drupal-10-will-be-supported-until-the-releas... → there will still be Drupal 10.3, 10.4 and 10.5, and PHPUnit 10/11 support is something that would be backported if the extent of the allowed changes allow it, because it supports contributed projects to be compatible with multiple Drupal major version. Is it not possible to target this at Drupal 10 even in that light? Especially of question I think is Drupal 10.3 which will be in beta in a little over 3 weeks from now per https://www.drupal.org/about/core/policies/core-release-cycles/schedule#... →

Re #123 honestly, I have doubts it would be possible to backport, given the amplitude of the changes from PHPUnit 9 to PHPUnit 10.

I still see at risk to close all the open issues by D11, frankly. Then, another significant effort will be needed for supporting PHPUnit 11: https://github.com/sebastianbergmann/phpunit/milestone/46?closed=1. And we will be late at that again, with PHPUnit 12 https://github.com/sebastianbergmann/phpunit/milestone/52 being 3 to 6 months away at that point.

PHPUnit major releases timeline, being yearly, is unforgiving for frameworks like Drupal that is much more focused on keeping BC, unfortunately. It's a philosophical question, at the end.

Eager to read others' point of view here.

I think the most important thing is whether modules can write phpunit 10 ( and eventually 11) compatible tests against Drupal 10, so that they don't need to maintain two branches. The actual version of phpunit that those tests run against is secondary.

Beyond that, it would be nice if we could widen Drupal 10's phpunit constraint to allow phpunit 10, but seems like that will be useless if we have a hard requirement on symfony phpunit bridge. That then brings up whether we want to do some Drupal 10 surgery to make symfony phpunit bridge an optional requirement (fork it? put it behind a metapackage so it's skipped when phpunit 11 is installed?), I don't think it's worth spending time on that until we know what we're doing in 11.x fully, since that'll determine exactly how much gap needs to be bridged.

Actually updating to phpunit 10 and dropping phpunit 9 support seems out of the question due to the scale of changes we're having to do in core.

So far, nearly all the changes we've been committing to core for phpunit 10 have been to stop using deprecated features, and we've been able to backport those to Drupal 10 without any issues, and there is nothing stopping contrib modules from making those same changes. But error and deprecation handling are still @todo.

The big issue with phpunit 11, is will it be dropping support for phpunit 9 features that were only added in phpunit 10? If so, then Drupal 10 will be stuck unless we try to add forwards compatibility shims, this is also too far beyond where we currently are to worry about now I think.

Then the last issue is whether phpunit 9 is going to support the php versions that Drupal 10 is going to need to support, that's also going to be a pain if it doesn't, although it's unlikely to be the only dependency with that problem.

> if we could widen Drupal 10's phpunit constraint to allow phpunit 10

I just don't think this will be feasible. The scope of the rewrites in PHPUnit 10 is just too large to try and support both at once. However, most tests should work just fine in both versions, unless they use some deprecated features - but even in those cases the fixes are backportable if needed.

> The big issue with phpunit 11, is will it be dropping support for phpunit 9 features that were only added in phpunit 10?

PHPUnit 11.0 is already out. PHPUnit works by soft-deprecating in one version, hard-deprecating in the next, and removing in the one after, so anything soft-deprecated in 9 will be removed in 11, but we should already have dealt with most of those as we have been on PHPUnit 9 for a long time now. https://phpunit.de/announcements/phpunit-11.html has a list of the breaking changes and some things are going away in PHPUnit 12, but we don't have to worry about that for a while at least.

PHPUnit 8 and 9 will continue to receive PHP version upgrades as long as their team decide it is feasible, so we can only hope that it will last for the lifetime of Drupal 10, but there are no guarantees. Symfony 6 will presumably have the same issue, as we share the same lifecycle.

PHPUnit 11.0 is already out. PHPUnit works by soft-deprecating in one version, hard-deprecating in the next, and removing in the one after, so anything soft-deprecated in 9 will be removed in 11

Forgot that detail - that's a good point so we should be fine split between phpunit 9 and 10 and just recommend people deal with all phpunit 9 deprecations.

Few updates to the IS.

We should update PHPUnit as https://github.com/sebastianbergmann/resource-operations is also an archived library and is required only on the last version supported by Drupal 9.6.17.

+1

mondrake → closed merge request !2464

We're down to the 2 last issues to be committed, 📌 Remove DrupalComponentTestListenerTrait and replace with a PHPStan rule RTBC and 📌 Upgrade PHPUnit to 10, drop Symfony PHPUnit-bridge dependency Fixed , then this one would be done - 3 years and 160 issues later :)

PHPUnit 10 is done! I wonder if we want to keep this open to track issues with the change record and/or discovered by contrib modules once thy start testing against it (or a new issue to track that?) otherwise I think we can close this too.

Seems like this just broke the current D11 (next major) Gitlab template. See e.g. https://git.drupalcode.org/issue/htmx-3447092/-/jobs/1591833

The problem is the use of the --no-interaction CLI flag, which no longer exists in 10.5.20

@fgm see this for the template adjustments: 📌 Prepare for PHPUnit 10 Fixed .

This looks done to me. Feel free to reopen if we want somewhere to track stuff, but moving it away from RTBC to reflect state of queue.

Automatically closed - issue fixed for 2 weeks with no activity.