KillTimer needs protection "if HandleAllocated then"

[User4martin]

KillTimer requires a handle.

However it is called in places that may be triggered even if there is no handle yet (and the VTV or one of its parents may not even have a parent themself, and handle creation may be impossible)

Example

procedure TBaseVirtualTree.DoChange(Node: PVirtualNode);

begin
  KillTimer(Handle, ChangeTimer);

Create a VTV in code, but do not set a parent yet, then trigger a change.

---

I propose to protect all KillTimer calls. Introducing a new Method KillTime(ID: ...). Checking for HandleAllocated can then be done within this, and the call with the handle can be done if safe.

---

For reference, the issue was discovered here https://gitlab.com/freepascal.org/lazarus/lazarus/-/issues/41430

I have not tested it with this VTV here, but a glance at the code made me believe that it can happen with this VTV too.

---

For reference / if interested in syncing with related other VTV
https://github.com/TurboPack/VirtualTreeView/blob/master/Source/VirtualTrees.pas#L7039-L7044
has StopTimer doing exactly this.

[User4martin]

Not entirely sure, but it may also apply to some "SetTimer" calls

[User4martin]

And maybe also calls to Handle in

• TBaseVirtualTree.DoSetOffsetXY
• InvalidateColumn (and similar)
• ToggleNode
• ClearSelection
• UpdateVerticalScrollBar (called eg from AddChild)

Just from a quick scan. There may be others

---

Also DestroyHandle kills two of the many timers, but not the others? Intention? Does the OS not kill those timers?

[User4martin]

I made a PR #36 for the KillTimer part. The remainder needs to be addressed separately.

I don't have any apps of my own using this VTV. So I can't do much testing beyond compilation and review. I am testing the same kind of changeset against the Lazarus included VTV, and so far no issues there.