Skip to content

Conversation

@luciansmith
Copy link
Member

libsbml is leaking memory in the loadPlugins functions, and my working theory is that this is because some constructors call it multiple times: once in the base class, then again in the derived class. This change removes all calls of loadPlugins for all derived classes where the base class also calls loadPlugins.

I'm not sure why we don't call loadPlugins in the SBase constructor directly, but changing that would be a fairly significant design change, so no need to try it now.

I believe that we cannot do the same thing with 'connectToChild', because it becomes important when it's the derived class that overrides the base definition for it, and I think that a base class constructor would not call the derived class version. If I'm wrong, we can revisit, but it's also true that connectToChild doesn't leak memory, so it's not dangerous to call it multiple times.

'setElementNamespace' also only needs to be called once, but this is sort of incidental to the other cleanups.

The change didn't end up fixing the leak, but it's still more efficient.

libsbml is leaking memory in the loadPlugins functions, and my working theory is that this is because some constructors call it multiple times: once in the base class, then again in the derived class.  This change removes *all* calls of loadPlugins for all derived classes where the base class also calls loadPlugins.

I'm not sure why we don't call loadPlugins in the SBase constructor directly, but changing that would be a fairly significant design change, so no need to try it now.

I believe that we cannot do the same thing with 'connectToChild', because it becomes important when it's the derived class that overrides the base definition for it, and I *think* that a base class constructor would not call the derived class version.  If I'm wrong, we can revisit, but it's also true that connectToChild doesn't leak memory, so it's not dangerous to call it multiple times.

'setElementNamespace' also only needs to be called once, but this is sort of incidental to the other cleanups.

I have not yet tested whether this change actually fixes the leak--needed to check this in from my development machine so I could check it out on my valgrind-enabled machine.
@luciansmith
Copy link
Member Author

OK, I went through the changes, fixed a bug, and I think it's ready to go. There are probably similar fixes that could be made in other packages, but this is good for now!

The basic idea is that we don't need to call setElementNamespace or loadPlugins in constructors where those functions are already called in their base classes, but you do still want to call connectToChild(), since that functions on sub-class elements.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant