dev/drupal#176 - Allow wider range of psr-log versions to be installed

[demeritcowboy]

Overview

https://lab.civicrm.org/dev/drupal/-/issues/176

This is an alternate version of (ubuntu 22.04-related) #23297

Before

Only psr-log 1.x allowed

After

More

Technical Details

Note that composer.lock is not updated, so this only changes the actual installed version for Drupal 10 and customized builds. Drupal 9 uses 1.1.4 (or at most ~1.0 if not using core-recommended). Drupal 10 requires 2.0, and it sounds like Ubuntu 22.04 requires 3.0 for the bundled civicrm package.

Comments

[civibot]

(Standard links)

• If this is your first pull-request for CiviCRM, please browse CONTRIBUTING.md for information about the development and testing processes.
• If you are reviewing this pull-request, you may wish to consult the test sites and the Review Standards (long template, short template).

[totten]

Doesn't it need a similar update for all of these files?

$ grep -r 'extend.*Logger' Civi CRM/ ext api/ tests/
Civi/Api4/Action/System/RotateKey.php:    $logger = new class() extends \Psr\Log\AbstractLogger {
CRM/Core/Error/Log.php:class CRM_Core_Error_Log extends \Psr\Log\AbstractLogger {
CRM/Utils/EchoLogger.php:class CRM_Utils_EchoLogger extends Psr\Log\AbstractLogger implements \Psr\Log\LoggerInterface {
CRM/Utils/SystemLogger.php:class CRM_Utils_SystemLogger extends Psr\Log\AbstractLogger implements \Psr\Log\LoggerInterface {

[demeritcowboy]

I didn't know those things existed. Civi has so much stuff.

[totten]

Revision looks good.

Yesterday, I ran the regular test-suite with psr/log v3 by unlocking the dependencies (ie rm composer.lock) (and bumping the metadata in a related package). The basic process for testing was like this:

civibuild download tmp-1-1 --type drupal-clean --patch https://github.com/civicrm/civicrm-core/pull/23409
cd tmp-1-1/web/sites/all/modules/civicrm

composer config platform.php 8.0
rm -rf composer.lock  vendor/
composer install

civibuild reinstall tmp-1-1
civi-test-run -b tmp-1-1 -j /tmp/junit all

All the substantive tests passed. (Formally, there was a failure in CRM_Core_ComposerConfigTest::testHardLocks(), but that just means it ran unlocked.)

I didn't do a full test-run with psr/log v2, but that doesn't worry me since (a) v1+v3 both pass and (b) this signature worked for v1+v2+v3 fine in a previous synthetic experiment.

Going forward, I think we should probably have a test-job which does something like this (ie run all tests, but with unlocked dependencies).

[demeritcowboy]

Thanks!