Remove class-by-value support in DataStorm #3251
Conversation
| @@ -240,58 +240,6 @@ namespace DataStorm | |||
| static T decode(const Ice::CommunicatorPtr& communicator, const Ice::ByteSeq& value) noexcept; | |||
| }; | |||
|
|
|||
| /** | |||
| * The Cloner template provides a method to clone user types. | |||
There was a problem hiding this comment.
Removed Cloner as it was used/tested only for "class by value". Its documentation (in the IceStorm manual) is also incomplete and rather confusing/incorrect.
There was a problem hiding this comment.
According to the documentation, the cloner template is also used by partial updates for cloning types. I believe it is required for class-by-reference scenarios to ensure ice_clone is called instead of using the shared_ptr copy constructor.
It might be worth updating DataStorm/partial to test class-by-reference behavior, as the current tests only cover structs. The original implementation tested a class-by-value, which is a removed by this PR.
From my understanding, without this change, the partial update will modify the previous value directly rather than computing a new value from previous.
| }; | ||
|
|
||
| /** | ||
| * Encoder template specialization to encode Ice::Value instances. |
There was a problem hiding this comment.
Only used for class-by-value.
This PR removes support for using Slice classes "by value" as DataStorm values.
You can still use Slice classes as reference with DataStorm; if you're currently using Slice classes "by value", you can switch to "by reference" and maintain on-the-wire interop with earlier DataStorm/Ice releases.