Remove Variable subsystem

Why are we opening a new critical task... Can't we do it in the meta? There was a patch in that issue already provided.

A meta issue is not supposed to perform functional changes and the actual functional change might require more discussion.

I'm not sure that this actually reflects current practice, but I'm entirely on board with marking #1775842: [meta] Convert all variables to state and/or config systems fixed when the blockers sun lists are fixed, and then handling the actual removal in this new issue. And this is critical, and a beta blocker. Thanks @sun.

Tag fail.

We also need to remove the variable table itself, and @berdir pointed out that "there's also a VariableTest that you can remove and SystemTestController::variable_get() including the route that points to it"

There are still some direct uses of the variable table in update-related functions, but these can be ignored because these functions are now only useful as a reference for the migrate initiative.

There may also be places looking directly into the global $conf variable for items that used to be variables. The $conf variable is still used by CMI though - see #1881582: Change configuration overrides to use $config instead of $conf

Yes, database drivers do not and cannot depend on the variable table, feel free to nuke this. At one point PostgreSQL depended on the schema, but we nuked that a long time ago.

There are also a number of global $conf of which most can be removed now, and we have #1881582: Change configuration overrides to use $config instead of $conf for the remaining ones.

Suppose statistics.php could skip page cache #2094335: statistics.php needs a lower bootstrap level

Re-roll.

I was able to bootstrap with this now, so I think it's time to let the testbot have a go with this :) Language negotiation tests will obviously still fail, and there's a weird node access test variable_set() left-over, but the issue for the first part is already RTBC and I want to review and RTBC #2175739: Actually test node_access_test_secret_catalan and clean up deprecated code in NodeAccessLanguageTest asap.

Uhm, no idea where that was coming from.

Just committed the negotiation CMI conversion so this ought to pass now.

Blocked by #2175739: Actually test node_access_test_secret_catalan and clean up deprecated code in NodeAccessLanguageTest

And also blocked by #2182603: Remove LocaleUninstallTest

Re-roll, Removed $new_phase, the second call has been removed already.

Fixed a notice in WebTestBase, that stuff there might need some more work...

Cleaning up DUBT tests.

FYI: #2175739: Actually test node_access_test_secret_catalan and clean up deprecated code in NodeAccessLanguageTest and #2182603: Remove LocaleUninstallTest are both committed as of moments ago. :)

Reroll since core/lib/Drupal/Core/EventSubscriber/ConfigGlobalOverrideSubscriber.php no longer exists and a couple of minor comment edits since the phrase "Drupal variables" doesn't really make any sense.

Improving issue summary and adding a suggested commit message.

+++ b/core/includes/bootstrap.inc
@@ -139,19 +139,14 @@
+const DRUPAL_BOOTSTRAP_FULL = 4;

And then there were 4!

Other than hook_update_N, upgrade tests, Migrate code, and core/scripts/generate-d6-content.sh, all variable code is gone.
Most of the additions to this patch are working around the setting of the request, or adding @todos.

I don't see why this isn't RTBC, so marking as such.

There might still be a few test fails that need to be looked into, if you look at the test fails in #20, there is for example MigrateSimpletestConfigsTest that fails and that does look suspicios...

I've been out of things for a while, but what an amazing accomplishment! Three years of work finally coming to fruition. Can't thank everyone who has helped out on CMI enough.

While we're waiting on testbot, would someone be able to take on crafting an appropriate commit message that at least attempts to credit all of the various people involved? :)

We still need to initialise global $conf in drupal_settings_initialize() since this is where settings.php is included.

Here are the 80 people, in alphabetical order, who were credited on a patch with the commit message "Convert ... configuration system" or "Convert ... CMI". Sorry if I missed anyone.

ACF, acrollet, adamdicarlo, Albert Volkman, alexpott, andreiashu, andyceo, andypost, anenkov, aspilicious, barbun, beejeebus, berdir, boombatower, cam8001, chriscalip, chx, cosmicdreams, dagmar, damiankloip, dawehner, deviance, disasm, dixon_, dstol, ebrowet, Gábor Hojtsy, heyrocker, Hydra, ianthomas_uk, japicoder, jcisio, jibran, julien, justafish, jvns, KarenS, kbasarab, kim.pepper, larowlan, Lars Toomre, leschekfm, Letharion, LinL, lirantal, Lukas von Blarer, marcingy, Mike Wacker, mrf, mtift, mtunay, n3or, nadavoid, nick_schuch, Niklas Fiekas, no_commit_credit, ParisLiakos, pcambra, penyaskito, pfrenssen, plopesc, Pol, Rok Žlender, rvilar, sun, swentel, tim.plunkett, tobiasb, tsvenson, typhonius, vasi1186, vijaycs85, wamilton, webchick, webflo, wizonesolutions, xjm, yched, YesCT, znerol

Bad testbot, bad! :)

Messing with our tests, tsss...

Will provide another patch that removes all the $conf/$config related todo's, I think that's unecessary, will just unecessary conflict with patches that are already in that issue and once this is in, it will simply have to get rid of all the $conf left-overs.

Here's a patch removing all the @todos

Here's another patch that does the same and changes/reverts a few more things.

- Revert the changed variable order in drupal_settings_initialize()
- Revert the removal of the similar global $conf in _drupal_load_test_overrides(), I think we should continue to support config overrides like that and will just have to rename it then.
- Revert changes to the update_variable_*() function docblocks, they will be removed in #2168011: Remove all 7.x to 8.x update hooks and disallow updates from the previous major version, which is close to RTBC as well, so let's not mess with that unnecessarily.
- Instead of simply dropping it, converted the $conf removal in DrupalKernelTest to settings, which is actually what the PhpStorageFactory is now using. There was also a second usage of $global conf in there which I updated too
- Instead of simply removing the bogus DatabaseStorageTest $conf change, I replaced it with a setting, to make it explicit that we are using the database backend and aren't just relying on the default.

I think we need one additional follow-up to move container_service_providers to settings instead of using global $conf, we have one for $config and for system.file:path.private.

Opened #2183323: Replace $GLOBALS['conf']['container_service_providers'] in DrupalKernel with Settings

Arg, copy & paste fail.

42: variable-die-2167109-42.patch queued for re-testing.

Edit: Random fail.

Committed/pushed to 8.x, thanks!

http://drupalcode.org/project/drupal.git/commit/e373ebe

Committed e373ebe and pushed to 8.x. Thanks!

A change notice already exists though it needs a lot of work :) https://drupal.org/node/1785368

Correct change notice: https://drupal.org/node/1500260

I've updated https://drupal.org/node/1500260/revisions/view/6872993/6873033 a tiny bit for changes in core to code style and the typification of config yaml, but not sure what else would be needed to cover this issue. That already mentions variable_get()/set/del are replaced with the config() API. Maybe more examples would be nice, eg. on how to del/remove things. But otherwise what is missing?

Added change notice for the actual change in https://drupal.org/node/2183531 - this was confirmed to be looking good by @beejeebus and @berdir, so closing this issue. Yayayayay!