Clang static analyzer falsely reports "1 uninitialized field at the end of the constructor call"

[jimis]

In general we get plenty of false positives, but here the logic looks simple enough for the analyzer to be able to track it. Details follow:

Detected with CodeChecker 6.18.2
Clang version: 13.0.1
Checker name: optin.cplusplus.UninitializedObject
Checker message: 1 uninitialized field at the end of the constructor call

Attaching the issue as a screenshot from CodeChecker as it is easier to follow the logic.

The code is at: https://github.com/qt/qtbase/blob/dfb4697e4a/src/corelib/text/qstring.h#L1602-L1647

[llvmbot]

@llvm/issue-subscribers-clang-static-analyzer

[haoNoQ]

@Szelethus do you think this is a checker bug or something more fishy is going on, like binding not showing up in RegionStore, or even if it does we're unable to load it?

[Szelethus]

At first glace, I'd say its the latter. The checker only emits reports if it finds an UndefinedVal.

[steakhal]

Likely fixed by 39743580aff515ba47cb7f9ffb2b5c538cff91da
I can't prove that it's really fixed due to the lacking reproducer, but I'm highly confident that this was also affected by it by looking at the curlyies in the init-list expression.

[jimis]

Thank you! Which clang version is the fix expected to be in?

[steakhal]

Thank you! Which clang version is the fix expected to be in?

In clang-18 around February, unless you take this commit and cherry-pick it to a clang-17, release and build it yourself.

[Snape3058]

A test case has been added covering this issue in #71073