[bugfix] I/O processing procedure: fixed bug (memory leakage) #49
Conversation
phlndrwd
left a comment
phlndrwd
left a comment
There was a problem hiding this comment.
Good spot, Lorenzo! There's me thinking that smart pointers prevented memory leaks. This wouldn't have happened with good old new and delete.
Well, I don't know ... I am not sure about that. P.S. -- C++ ... a such (very) easy, wonderful language ... |
|
@odlomax I think that @mo-lormi's work here is an interesting commentary on the pitfalls of smart pointers. I understand we're using them because of the perception that they prevent memory leaks, but this story shows how they introduce another set of rules and procedures that need to be understood to prevent them.
If I were writing it. I learned |
Description
This PR is intended to fix a bug (memory leakage) in the I/O processing procedure part of the MONIO package
Recently, debugging/profiling of the application
lfricjedi_hofx.x(OOPS class:oops::HofX4D) has been carried out by using the AddressSanitizer (ASan) tool.During this investigation, a few memory leaks have been identified.
A subset of them are associated with the I/O processing procedure part of the MONIO package.
* debugging/profiling (before)
The output of the debugging/profiling can be found here.
* solution:
The memory leaks occurs because the resources allocated for the data structure
dataFile_are not correctly deallocated at the end.The bug/issue is related with the call of the destructor of
std::unique_ptr, in particular it is related with the difference between the two methodsstd::unique_ptr::release()andstd::unique_ptr::reset().For details see here, here, and here.
* debugging/profiling (after)
The output of the debugging/profiling carried out after fixing the bug can be found here.
Issue(s) addressed
issue #48
Acceptance Criteria (Definition of Done)
All the tests pass.