https://bugs.documentfoundation.org/show_bug.cgi?id=170805

Michael Weghorn <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Regression By|                            |Michael Weghorn
           Assignee|[email protected] |[email protected]
                   |desktop.org                 |
             Status|NEW                         |ASSIGNED
           Keywords|                            |bibisected, bisected,
                   |                            |regression

--- Comment #4 from Michael Weghorn <[email protected]> ---
(In reply to Mike Kaganski from comment #3)
> Most likely regression after 3c20c8ed01bde299d3e40186edcd3a6021baeee3 - hard
> to tell because bisect repos lack automatic update check.
> 
> No, it's not "always non-null": it is destroyed in the call to
> myExtMgr->Close() in ServiceImpl::startExecuteModal, and the thread happily
> tries to dereference the destroyed object.

Thanks for the analysis.

I don't see any issue with 3c20c8ed01bde299d3e40186edcd3a6021baeee3 itself. 
`Thread::m_pDialogHelper` was technically "always non-null" as it was never set
to null before the above commit. (It was non-null, but a dangling pointer after
the object was destroyed before the above commit, i.e. the problem was
effectively the same as the reference to the deleted object with the commit in
place.)

This is a regression from another commit part of some extension manager
refactoring, however:

    commit 5a6985c124c7a0ca246fbc920ab5e024d72a5e4a
    Author: Michael Weghorn
    Date:   Sat Nov 8 09:42:17 2025 +0100

        tdf#169318 extension mgr: Deduplicate close logic

Pending fix:

https://gerrit.libreoffice.org/c/core/+/200151

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to