- ensure the prefix is changed when an app is (dis)enabled
- ensure the app ids are included in the hash instead of only the
version numbers
- ensure hash is deterministic by always use the same order.
Signed-off-by: Ferdinand Thiessen <opensource@fthiessen.de>
This removes a circular dependency between AppConfig and cache factory.
When a cache in the app config is used.
Signed-off-by: Ferdinand Thiessen <opensource@fthiessen.de>
but fallback to NullCache. This can be the case if APCu is used without apc.enable_cli=1. APCu however runs within the PHP instance and hence cannot be shared between CLI and web or used as distributed cache. The CLI call can hence only create or invalidate entries for itself. For short-living CLI calls, this is theoretically a downsides regarding performance and resource usage, and Nextcloud must not have any issues using the dummy NullCache instead of an isolated freshly created and destroyed APCu instance.
This partly reverts https://github.com/nextcloud/server/pull/25770. The fallback was removed, because CLI calls started to hang after https://github.com/nextcloud/server/pull/25440. The reason however was not that a cache is generally required for CLI calls, but because the previously logged warning invoked the user backend, which invoked again the caching backend, causing a loop.
This commit re-adds the fallback without logging a warning, and for APCu only. For mentioned reasons, it is okay to fallback to NullCache silently. If Redis or memcached are configured but not available, then the web UI would fail as well, and it makes sense to abort and inform CLI calls as well then.
The motivation is to make apc.enable_cli=1 optional, and that hence the documentation about it can be removed. We should not enforce admins to enable APCu for CLI calls, which is reasonably disabled by default. This also reduces requirements for hosting providers to support Nextcloud.
Signed-off-by: MichaIng <micha@dietpi.com>
The intention obviously was to check whether $lockingCacheClass is defined, and existing class, and available. It was however checked whether the $distributedCacheClass is an existing class, which would have caused an exception already in the previous distributed cache check.
Signed-off-by: MichaIng <micha@dietpi.com>
To continue this formatting madness, here's a tiny patch that adds
unified formatting for control structures like if and loops as well as
classes, their methods and anonymous functions. This basically forces
the constructs to start on the same line. This is not exactly what PSR2
wants, but I think we can have a few exceptions with "our" style. The
starting of braces on the same line is pracrically standard for our
code.
This also removes and empty lines from method/function bodies at the
beginning and end.
Signed-off-by: Christoph Wurst <christoph@winzerhof-wurst.at>
