How hard would it be to change to QScopedPointer?

Hmm, I see your point. To put this change in context: I have run KDevelop inside Valgrind since I wanted to debug a crash-on-close problem. Valgrind however showed tons of incomplete deletions of LanguageSupport objects, which vanished after adding these explicit deletes.
Do you think that these Valgrind errors may be caused by having a crash in KDevelop before the QObject children relationships could take care for deleting the m_highlighting and m_refactoring objects?

In D1555#29111, @cordlandwehr wrote:

Hmm, I see your point. To put this change in context: I have run KDevelop inside Valgrind since I wanted to debug a crash-on-close problem. Valgrind however showed tons of incomplete deletions of LanguageSupport objects, which vanished after adding these explicit deletes.

I see.

Do you think that these Valgrind errors may be caused by having a crash in KDevelop before the QObject children relationships could take care for deleting the m_highlighting and m_refactoring objects?

Likely, yes. I've been running KDevelop under ASAN for a while (with leak reports enabled), and I fixed tons of leaks already. I would have seen this one. You agree that this deletion should be taken care of via QObject relationship, right? In that case, we can abandon this patch.