- Issue created by @Liam Morland
- π¨π¦Canada joseph.olstad
D7 default should now be set to PHP 8.2
AND
PHP 8.3 automated tests should be added here: - πΈπ°Slovakia poker10
Thanks for creating the issue. Let's try how PHP 8.3 test suite runs on D7 core currently. Adding a noop patch.
- last update
10 months ago Patch Failed to Apply - last update
10 months ago Patch Failed to Apply - last update
10 months ago Patch Failed to Apply - last update
10 months ago Patch Failed to Apply - last update
10 months ago 2,158 pass, 3 fail - last update
10 months ago 2,119 pass, 3 fail - last update
10 months ago 2,154 pass, 3 fail - πΈπ°Slovakia poker10
This looks good. There seems to be only one problem with unserialization in D7 core tests (there could be more, but not catched by tests).
I have created an issue with the fix for the unserialize failure: π [D7 PHP 8.3] unserialize(): Extra data starting at offset Fixed
- πΈπ°Slovakia poker10
The child issue is fixed. Let's check with another noop patch if we are green now.
- last update
8 months ago 2,155 pass, 2 fail - last update
8 months ago 2,119 pass, 3 fail 47:03 46:16 Running47:03 46:41 Running47:03 46:16 Running47:03 46:42 Running- last update
8 months ago 2,157 pass - last update
8 months ago 2,161 pass - last update
8 months ago 2,122 pass - last update
8 months ago run-tests.sh exception - πΈπ°Slovakia poker10
Seems like other failure/s appeared in PHP8.3-RC tests, so we need to check these as well (probably something must have changed between PHP8.3 beta and RC, because we were already green after #3380876).
- πΈπ°Slovakia poker10
Seems like the
TextSummaryTestCase
failures could be related to similar failures in D10: π TextSummaryTest:testLength() fails on some libxml versions Closed: outdated . So we need to create an issue for this in D7 as well. - πΈπ°Slovakia poker10
Created new child issue π [D7 PHP 8.3] TextSummaryTestCase::testLength() fails on some libxml versions Active .
- last update
7 months ago 2,159 pass, 2 fail - π«π·France andypost
queued last patch on 8.3.0RC5 looks the only broken test!
- π¨π¦Canada joseph.olstad
Nice, only 2 fails! Subscribing.
PHP 8.3 is scheduled to release on November 23, 2023
Release candidate 6 is in a few days from now, perhaps we get lucky with rc6 passing more tests than rc5?
or is this a documented/expected fail?
- πΈπ°Slovakia poker10
I do not think that the failures will disappear, as it seems like that they are caused by some changes in libxml library, not directly by PHP 8.3 (see the child issue).
- last update
7 months ago 2,161 pass, 2 fail - last update
7 months ago 2,161 pass, 2 fail - π«π·France andypost
Failed test for session needs new issue and still active π [D7 PHP 8.3] TextSummaryTestCase::testLength() fails on some libxml versions Active
- Merge request !8020Draft: Issue #3380823: [META] Make Drupal 7 core compatible with PHP 8.3 β (Open) created by poker10
- πΈπ°Slovakia poker10
I have created another child issue for the session failure.
testEmptySessionId() failure: π [D7 PHP 8.3] Fix SessionHttpsTestCase->testEmptySessionId() failure Needs review
testLength() failure: π [D7 PHP 8.3] TextSummaryTestCase::testLength() fails on some libxml versions Active
Created also a draft MR here with PHP 8.3 testing (https://git.drupalcode.org/project/drupal/-/merge_requests/8020) and after applying fixes from both issues, tests are green, see: https://git.drupalcode.org/project/drupal/-/pipelines/170105 (sorry for multiple commits, but I had/have issues with replicating the session failure locally).
- πΈπ°Slovakia poker10
Hiding the old patch as we have now a MR here for testing.
- π¨π¦Canada joseph.olstad
@poker10, great work!
So if I understand correctly, you've put all the fixes into the MR 8020, then split off the actual code that does the fixes into those two issues 3446569 and 3397882 ?
Right now I see 100% green in the MR 8020 - πΈπ°Slovakia poker10
Yes - both fixes are in MR 8020 to see the final effect (but this MR will not be committed) and then there are two separate issues with MRs to commit.
I have done it because the MR workflow is more flexible and also because I was not able to simulate the session test failure locally, so I needed to experiment a bit.