RepositoryResolver.Current being initialized which normally doesn't occur until after the ServiceContext
is constructed. Adds instance level caching for the GetRecursiveValue method in case this is called more than
one time for a property in one view. Reverts PetaPocoUnitOfWork to allow more than one call to Commit().. this
isn't 'best practices' per se but it is more for performance reasons because otherwise we'd have to create a new
repo object + uow for any bulk saving operations... The end result is the same, bulk operations in the Services
cannot be processed in one transaction. Fixing up the ContentServiceTests by ensuring that the shared and always open sqlce
connection with the legacy SqlCeContextGuardian is closed on TearDown.
Modified the CE one so that it is consistent with the Sql one, we should be relying on the same scripts here (or at least the same logic)
the one for CE will not work because even if the db is not installed it would have still returned 0 which would pass and the installer would have
thought it was installed.
so have left it's installer flagged as 4.8 (just means that the intall will just hang... not sure why we have this flag anyways??
shouldn't we always support the latest version for all dbs??)
Had to fix the ApplicationTree and Application classes to lazily call the Cache method
(which is no renamed to EnsureCache) so that we could set the config location in the unit tests.
This should also improve performance a teeny bit and allow for better error handling if required.
Updated InternalsVisibleTo declarations.
Updated the ApplicationTest test to use a new base class as a test to install a local sqlce database
for testing against.
Added Umbraco.Core project.
Added Umbraco.Web project.
Created new sln folder Legacy
locally when debugging. Updated build script to not have to manually build the umbracosqlce project
since its now referenced.
Updated ignore file.
Removed Umbraco.sdf as being referenced in the repository.
Fixed SqlCEInstaller to automatically create the sqlce database if it doesn't exist, we now don't ship the empty database
and this is much nicer for working in the core.
