Follow-up from "Coverity issues of type "resource leak"."
The following discussions from !347 (closed) should be addressed:
-
@graven started a discussion: (+2 comments) Instead of fixing the specific symptoms, why not be rigorous, and use
std::unique_ptr
instead? (which requires adapting a few interfaces, but that actually clarifies what those interfaces are expected to do!) -
@graven started a discussion: (+4 comments) why not use an
std::unique_ptr
with a custom deleter which callsclose
in its destructor? That fixes the origin of the problem instead of fixing the symptom... -
also include the only non controversial change in !347 (closed)