Libconverter calls Importer::Load multiple times on linux systems to import different models, without freeing up the memory of the last loaded model

If you call libconverter’s Importer::Load multiple times on windows to import different models, the console memory print will automatically release the memory of the last loaded model and reload the new one, but under Linux, the memory will be accumulated if you load the model multiple times, is this a bug or a system mechanism?
windows


linux

Communicator version 2021_SP2_U2,I don’t know if this issue has been identified and fixed in subsequent versions

Hey Rui, I opened a request for you for this issue because I think it’s something that probably needs to be looked into a little more in-depth.