Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CRM-8597, CRM-20561 - Restore fix for & #3

Merged
merged 1 commit into from
Jun 16, 2017

Conversation

totten
Copy link

@totten totten commented Jun 16, 2017

No description provided.

@seamuslee001 seamuslee001 merged commit e737bf0 into seamuslee001:CRM-20561 Jun 16, 2017
seamuslee001 pushed a commit that referenced this pull request Sep 7, 2017
seamuslee001 pushed a commit that referenced this pull request Oct 29, 2018
…ing filters

Suppose you run a search ("Find Contact", "Advanced Search", "Custom Search", etc). The result screen includes
several elements (which we'll reference below):

1. Standard pagination (Previous/Next; First/Last; Jump-To)
2. Numerical option for page-size
3. Sortable columns
4. An alphabetical filter
5. Checkboxes

As you work with these options, the content of the `civicrm_prevnext_cache` table may change. This patch does
not substantively change what's in that cache, but makes the column `cacheKey` simpler and more consistent.

Both Before and After (Unchanged)
---------------------------------
* The form's qfKey identifies the current screen/filters/cache.
* If you navigate to the next/previous page (`#1`) or adjust the page-size (`#2`), the content in `civicrm_prevnext_cache` remains the same (for the given qfKey).
* If you change the sort column (`#3`) or alphabetic filter (`#4`), the content in `civicrm_prevnext_cache` is deleted and repopulated (for the given qfKey).
* If you toggle a checkbox, the `civicrm_prevnext_cache.is_selected` property updates accordingly. These selections are retained when changing pages (`#1`/`#2`),
  but they're reset if you use sort or alphabet options (`#3`/`#4`).

Before
------
* The content of `civicrm_prevnext_cache.cacheKey` takes one of two forms, depending on whether you're using an alphabetic filter (`#4`).
    * `civicrm search {qfKey}` (typical, without any alphabetic filter)
    * `civicrm search {qfKey}_alphabet` (less common, with an alphabetic filter)
* The queries which read or delete the query-cache use a prefix+wildcard, i.e. `WHERE cacheKey LIKE 'civicrm search {qfKey}%'`.

After
-----
* The content of `civicrm_prevnext_cache.cacheKey` takes only one form
    * `civicrm search {qfKey}`
* The queries which read or delete the query-cache use an exact match, i.e. `WHERE cacheKey = 'civicrm search {qfKey}'`.`
* The text `_alphabet` does not appear in the PHP source folders (CRM, Civi, bin, api, extern, tests).

Comments
--------
In theory, one can imagine that it's desireable to keep the cached results for each of the sorted/filtered variants of the query.
That might allow the user to quickly switch among different sortings and different alphabetic-filters, or it might
allow some kind of clever management of the selections. But this is not so. As we see (both before and after), the substance
of the cache is deleted whenever the user changes `#3`/`#4`. In reality, one user browsing a search screen corresponds to exactly
one query-cache. As near as I can tell, the old code made the names change for no real reason at all.

To observe the behavior empirically, I would twiddle the UI widgets and concurrently inspect the content of the cache tables.  For example:

```
mysql> select group_name, path, FROM_BASE64(data), expired_date  from civicrm_cache where path like 'civicrm search%';
select 'Total records in' as label, cacheKey, count(*), min(id), max(id) from civicrm_prevnext_cache group by cacheKey
union select 'Selected records in ', cacheKey, count(*), min(id), max(id) from civicrm_prevnext_cache where is_selected=1 group by cacheKey;
+------------------------------+------------------------------------------------------+--------------------------------------------------------------+--------------+
| group_name                   | path                                                 | FROM_BASE64(data)                                            | expired_date |
+------------------------------+------------------------------------------------------+--------------------------------------------------------------+--------------+
| CiviCRM Search PrevNextCache | civicrm search a8ed1e2039241c41457a88f65aa8a8ee_7845 | s:52:"civicrm search a8ed1e2039241c41457a88f65aa8a8ee_7845"; | NULL         |
+------------------------------+------------------------------------------------------+--------------------------------------------------------------+--------------+
1 row in set (0.00 sec)

+----------------------+------------------------------------------------------+----------+---------+---------+
| label                | cacheKey                                             | count(*) | min(id) | max(id) |
+----------------------+------------------------------------------------------+----------+---------+---------+
| Total records in     | civicrm search a8ed1e2039241c41457a88f65aa8a8ee_7845 |        6 |     787 |     792 |
| Selected records in  | civicrm search a8ed1e2039241c41457a88f65aa8a8ee_7845 |        1 |     789 |     789 |
+----------------------+------------------------------------------------------+----------+---------+---------+
2 rows in set (0.01 sec)
```
seamuslee001 pushed a commit that referenced this pull request Feb 19, 2020
Overview
--------

This fixes a recent regression in which `xml/GenCode.php` fails to execute in certain configurations.

Initial report: civicrm/civicrm-buildkit#503

Before
------

* If the folder `l10n` exists in its traditional location, and if you run `./bin/setup.sh -g`, then it
  correctly executes.
* If the folder `l10n` does not exist in its traditional locacation, and if you run `./bin/setup.sh -g`,
  then it fails with an error like this:
  ```
  + php -d mysql.default_host=127.0.0.1:3306 -d mysql.default_user=dmasterciv_abcde -d mysql.default_password=abcd1234 GenCode.php schema/Schema.xml '' Drupal

  civicrm_domain.version := 5.23.alpha1

  Parsing schema description schema/Schema.xml
  Extracting database information
  Extracting table information
  PHP Fatal error:  Uncaught RuntimeException: Invalid configuration: the [cms.root] path must be an absolute URL in /home/me/.../Civi/Core/Paths.php:174
  Stack trace:
  #0 /home/me/.../Civi/Core/Paths.php(151): Civi\Core\Paths->toAbsoluteUrl('/', 'cms.root')
  #1 /home/me/.../Civi/Core/Paths.php(176): Civi\Core\Paths->getVariable('cms.root', 'url')
  #2 /home/me/.../Civi/Core/Paths.php(151): Civi\Core\Paths->toAbsoluteUrl('/', 'civicrm.root')
  #3 /home/me/.../Civi/Core/Paths.php(224): Civi\Core\Paths->getVariable('civicrm.root', 'path')
  #4 /home/me/.../Civi/Core/Paths.php(84): Civi\Core\Paths->getPath('l10n')
  #5 [internal function]: Civi\Core\Paths->Civi\Core\{closure}()
  #6 /home/jo in /home/me/.../Civi/Core/Paths.php on line 174
  ```

After
-----

You can run `./bin/setup.sh -g` with or without the `l10n` folder.

Comments
--------

There's an argument to be made that this shouldn't be necessary: when running
`GenCode`, it should only create PHP files, and none of the output should depend
on the CMS URL - because that's liable to change when you deploy the PHP code.
If someone did try to generate URL at such an early stage, it's arguably good to
generate an error.  In point of fact, `GenCode` is not actually using the CMS
URL.

The oddity stems from the contract of `CRM_Utils_System_*` (specifically,
`getCiviSourceStorage()` and `getDefaultFileStorage()`) which do double-duty
providing both path and URL.  To avoid duplicate work, `Civi\Core\Paths` uses
the same grain of information - it tracks the pairs of path+URL.

A more aggressive fix might be to split `getCiviSourceStorage()` and
`getDefaultFileStorage()` so that it's possible to get the path and URL
separately; then revise `Civi\Core\Paths` to take advantage of the finer-grained
contract.  However, splitting those things would be a more invasive patch,
and we're currently in RC.
seamuslee001 pushed a commit that referenced this pull request Sep 4, 2020
…esources

This is a very subtle behavioral change.  To understand it, we must consider
the traditional ordering in `CRM_Core_Resources::addScriptUrl()` and
`CRM_Core_Region::add()`.  Compare these two examples:

```
Civi::resources()->addScriptUrl('http://example.com/foo.js', 100, 'page-footer');
Civi::resources()->addScriptUrl('http://example.com/bar.js', 100, 'page-footer');

CRM_Core_Region::instance('page-footer')->add([
  'scriptUrl' => 'http://example.com/foo.js',
  'weight' => 100,
]);
CRM_Core_Region::instance('page-footer')->add([
  'scriptUrl' => 'http://example.com/bar.js',
  'weight' => 100,
]);
```

You might expect this to output `<SCRIPT>`s in the same order.  After all,
the examples use the same URLs, the same weights, and the same
procedural/natural order.  Moreover, `Civi::resources()` is just a wrapper
for `CRM_Core_Region`.  Surely they were the same!

They were not. The output order would differ.

The reason is that the ordering had two keys (`weight,name`), and the
secondary-key (`name`) was hidden from us:

* In `CRM_Core_Resources::addScriptUrl()`, the default `name` was the URL.
  This was handy for de-duping resources.
* In `CRM_Core_Region::add()`, the default `name` was a numeric position.
  This made it behave like procedural/natural-ordering.

Of course, the goal in this branch is to make various collections support
the same methods and the same behaviors.  But they didn't agree before, so
something has to give.  Either:

1. Standardize on the behavior of `Resources::addScriptUrl()` and change `Region::add(scriptUrl)`.
2. Standardize on the behavior of `Region::add(scriptUrl)` and change `Resources::addScriptUrl()`.
3. Embrace a split. All `CRM_Core_Resources::add*()` behave one way, and all `CRM_Core_Region::add*()` behave another.
4. Embrace a split. All rich adders (`*::addScriptUrl()`) behave one way, and all generic adders (`*::add()`) behave another.

This developmental branch has been using `#3`. The patch changes it to `#4`.

Both splits achieve backward compatibility wrt `Resources::addScriptUrl()`
and `Region::add(scriptUrl)`.  The difference is that `#4` allows us to have
the same behavior in all collections (regardless of subtype), which means
that collections can be more interoperable and share more code.  From my
POV, it's still quirkier than `#1` or `#2`, but it's not as quirky as `#3`.
seamuslee001 pushed a commit that referenced this pull request May 18, 2021
Overview
--------

This slightly relaxes the default settings so that it is easier to use `authx` as a replacement for `extern/rest.php` (which
uses the `api_key` for authentication).

Before
------

`extern/rest.php` accepts `api_key`s.

`authx` can accept `api_key`s, but you have to change some settings to allow it.

After
-----

Both `extern/rest.php` and `authx` accept `api_key`s by default.

Comments
-----------------

The defaults in authx were designed to be a bit paranoid.  The basic goal -- don't let anyone open up access
accidentally.  However, the current protections are a bit of overkill.  Suppose you've created an `api_key` (sufficient
for `extern/rest.php`) and you want to switch to using `civicrm/ajax/*`. Here are the sysadmin tasks:

1. Enable `authx` (*It's inert otherwise.*)
2. Grant the user permission `authenticate via api key`, or convey the `SITE_KEY` to the user, or disable all `authx_guards`
3. Update the setting `authx_header_cred` or `authx_xheader_cred` or `authx_param_cred` to allow `api_key`

Notably, when this default was first chosen during drafting, the `authx_guards` (step 2) didn't exist. But now they do, and we have even more steps to go through.

This change relaxes the defaults so that step `#3` is not necessary. This arrangement is still fairly paranoid (ie we still have 1+2 -- compared
to `extern/rest.php`, there's still an extra opt-in hoop).
seamuslee001 pushed a commit that referenced this pull request Dec 7, 2021
…ion-fix)

Overview
--------

Fixes a recent regression that prevents you from uninstalling extensions via
CLI.  This specifically affects extensions which use managed entities.

Steps to reproduce
------------------

```
cv en afform
cv dis afform
cv ext:uninstall afform
```

Before
-------

```
[bknix-max:~/bknix/build/dmaster/web/sites/all/modules/civicrm] cv en afform && cv dis afform && cv ext:uninstall afform
Enabling extension "org.civicrm.afform"
Disabling extension "org.civicrm.afform"
Uninstalling extension "org.civicrm.afform"
Error: API Call Failed: Array
(
    [entity] => Extension
    [action] => uninstall
    [params] => Array
        (
            [keys] => Array
                (
                    [0] => org.civicrm.afform
                )

            [debug] => 1
            [version] => 3
        )

    [result] => Array
        (
            [error_code] => unauthorized
            [entity] => Extension
            [action] => uninstall
            [trace] => #0 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/Civi/API/Kernel.php(147): Civi\API\Kernel->authorize(Object(Civi\Api4\Provider\ActionObjectProvider), Object(Civi\Api4\Generic\DAODeleteAction))
 #1 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/Civi/Api4/Generic/AbstractAction.php(234): Civi\API\Kernel->runRequest(Object(Civi\Api4\Generic\DAODeleteAction))
 #2 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/api/api.php(85): Civi\Api4\Generic\AbstractAction->execute()
 #3 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/CRM/Core/ManagedEntities.php(467): civicrm_api4('OptionValue', 'delete', Array)
 #4 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/CRM/Core/ManagedEntities.php(303): CRM_Core_ManagedEntities->removeStaleEntity(Object(CRM_Core_DAO_Managed))
 #5 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/CRM/Core/ManagedEntities.php(134): CRM_Core_ManagedEntities->reconcileUnknownModules()
 #6 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/CRM/Core/Invoke.php(409): CRM_Core_ManagedEntities->reconcile()
 #7 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/CRM/Extension/Manager.php(483): CRM_Core_Invoke::rebuildMenuAndCaches(true)
 #8 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/api/v3/Extension.php(183): CRM_Extension_Manager->uninstall(Array)
 civicrm#9 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/Civi/API/Provider/MagicFunctionProvider.php(89): civicrm_api3_extension_uninstall(Array)
 civicrm#10 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/Civi/API/Kernel.php(149): Civi\API\Provider\MagicFunctionProvider->invoke(Array)
 civicrm#11 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/Civi/API/Kernel.php(81): Civi\API\Kernel->runRequest(Array)
 civicrm#12 /home/me/bknix/build/dmaster/web/sites/all/modules/civicrm/api/api.php(22): Civi\API\Kernel->runSafe('Extension', 'uninstall', Array)
 civicrm#13 phar:///home/me/bknix/bin/cv/src/Command/BaseCommand.php(49): civicrm_api('Extension', 'uninstall', Array)
 civicrm#14 phar:///home/me/bknix/bin/cv/src/Command/ExtensionUninstallCommand.php(63): Civi\Cv\Command\BaseCommand->callApiSuccess(Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput), 'Extension', 'uninstall', Array)
 civicrm#15 phar:///home/me/bknix/bin/cv/vendor/symfony/console/Command/Command.php(257): Civi\Cv\Command\ExtensionUninstallCommand->execute(Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput))
 civicrm#16 phar:///home/me/bknix/bin/cv/vendor/symfony/console/Application.php(850): Symfony\Component\Console\Command\Command->run(Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput))
 civicrm#17 phar:///home/me/bknix/bin/cv/vendor/symfony/console/Application.php(193): Symfony\Component\Console\Application->doRunCommand(Object(Civi\Cv\Command\ExtensionUninstallCommand), Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput))
 civicrm#18 phar:///home/me/bknix/bin/cv/src/Application.php(46): Symfony\Component\Console\Application->doRun(Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput))
 civicrm#19 phar:///home/me/bknix/bin/cv/vendor/symfony/console/Application.php(124): Civi\Cv\Application->doRun(Object(Symfony\Component\Console\Input\ArgvInput), Object(Symfony\Component\Console\Output\ConsoleOutput))
 civicrm#20 phar:///home/me/bknix/bin/cv/src/Application.php(15): Symfony\Component\Console\Application->run()
 civicrm#21 phar:///home/me/bknix/bin/cv/bin/cv(27): Civi\Cv\Application::main('phar:///Users/t...')
 civicrm#22 /home/me/bknix/bin/cv(14): require('phar:///Users/t...')
 civicrm#23 {main}
            [is_error] => 1
            [error_message] => Authorization failed
        )

)
```

After
-----

Works

Comment
-------

I encountered this while working on E2E test-coverage for other changes.
The E2E test coverage had worked on a previous iteration of 5.45.alpha1 but
failed when I rebased. Consequently, this means

You can see a prior draft of the E2E test [here](https://github.com/totten/shimmy/blob/master-reorg/shimmy/tests/phpunit/E2E/Shimmy/LifecycleTest.php#L56-L77).
However, it's being reworked as a core patch.

I'd suggest accepting this without a test - because (a) it's a regression and (b) there will be coverage from the pending change.
seamuslee001 pushed a commit that referenced this pull request Dec 8, 2022
preg_replace(): Passing null to parameter #3 () of type array|string is deprecated
totten pushed a commit that referenced this pull request Sep 19, 2024
CRM_Event_Page_EventInfoTest::testFullMessage
Exception: CRM_Extension_Exception_MissingException: "Failed to find extension: civi_mail"
#0 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Extension/Container/Basic.php(143): CRM_Extension_Container_Basic->getRelPath("civi_mail")
#1 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Extension/Mapper.php(233): CRM_Extension_Container_Basic->getPath("civi_mail")
#2 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Core/Resources.php(261): CRM_Extension_Mapper->keyToBasePath("civi_mail")
#3 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Core/Resources.php(311): CRM_Core_Resources->getPath("civi_mail")
#4 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/Civi/Angular/Manager.php(208): CRM_Core_Resources->glob("civi_mail", (Array:3))
#5 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/Civi/Angular/Manager.php(114): Civi\Angular\Manager->resolvePatterns((Array:59))
#6 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Utils/Check/Component/Env.php(1178): Civi\Angular\Manager->getModules()
#7 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Utils/Check/Component.php(76): CRM_Utils_Check_Component_Env->checkAngularModuleSettings(FALSE)
#8 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Utils/Check.php(215): CRM_Utils_Check_Component->checkAll((Array:0), FALSE)
civicrm#9 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Utils/Check.php(185): CRM_Utils_Check::checkStatus()
civicrm#10 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Utils/Check.php(93): CRM_Utils_Check::checkAll()
civicrm#11 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Core/Page.php(267): CRM_Utils_Check->showPeriodicAlerts()
civicrm#12 /home/homer/buildkit/build/build-3/web/sites/all/modules/civicrm/CRM/Event/Page/EventInfo.php(325): CRM_Core_Page->run()
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants