You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Until #1053 gets completed, the TestSystem is basically ignored in the integration with Typesafe Config.
Even though the external library does not support overriding of the System class method calls, there would be a way around this. Both system properties and the environment variables can be altered globally (the latter using a reflection hack), so the only remaining part is to assure, that all Typesafe Config reads are done sequentially - which should be possible using a JVM-global semaphore.
This would unblock testing of how do the environment variables resolve to the configuration values in complex cases.
Another advantage of the proposal is that it's safer for legacy apps (which use complex HOCON configurations already) since all the undocumented quirks of the 3rd party library would work as before. That would make it less risky to port them to ZIO.
The text was updated successfully, but these errors were encountered:
Both system properties and the environment variables can be altered globally (the latter using a reflection hack)
Since the hack will only work on certain JVMs, the code would need to check if using it is possible, and if not, then a descriptive exception should be thrown.
Until #1053 gets completed, the TestSystem is basically ignored in the integration with Typesafe Config.
Even though the external library does not support overriding of the System class method calls, there would be a way around this. Both system properties and the environment variables can be altered globally (the latter using a reflection hack), so the only remaining part is to assure, that all Typesafe Config reads are done sequentially - which should be possible using a JVM-global semaphore.
This would unblock testing of how do the environment variables resolve to the configuration values in complex cases.
Another advantage of the proposal is that it's safer for legacy apps (which use complex HOCON configurations already) since all the undocumented quirks of the 3rd party library would work as before. That would make it less risky to port them to ZIO.
The text was updated successfully, but these errors were encountered: