Bug#989293: unblock: breeze/4:5.20.5-4

[Norbert Preining <norbert@preining.info>]

Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: unblock
X-Debbugs-Cc: debian-qt-kde@lists.debian.org

Please unblock package breeze

[ Reason ]

The fix that was included in -3 introduced a different problem that was
recently fixed by upstream and pushed to stable branches.
KDE bug: 436473

The problem is that under certain circumstances the mouse cursor could
get stuck in form of resize icon.

Upstream comments are in the attached patch, I only include the
important paragraph here
> This was broken by accident in d201a1f187 ("Fix SplitterProxy not clearing
> when above another QSplitterHandle"), which moved the hide() call past the
> call to QCoreApplication::sendEvent. Previously that made isVisible() false,
> which also prevented the interception of the HoverLeave/HoverMove events.

[ Impact ]
If the bug remains unfixed, the mouse cursor might remain stuck in
resize form without a way to revert it.

[ Tests ]
Manual test trying to trigger the bug.

[ Risks ]
The code has been approved by upstream and included in 5.18 LTS and 5.20.

[ Checklist ]
  [x] all changes are documented in the d/changelog
  [x] I reviewed all changes and I approve them
  [x] attach debdiff against the package in testing


I attach only the patch that is added, the remaining changes are:
debian/changelog entry:
breeze (4:5.20.5-4) unstable; urgency=medium

  [ Norbert Preining ]
  * Upstream fix for cursor stuck in resize icon.

 -- Norbert Preining <norbert@preining.info>  Sun, 30 May 2021 05:58:28 +0900

and adding the patch to debian/patches/series.


unblock breeze/4:5.20.5-4

>From f99b7ef621c9c69544158d245699fd8104db6753 Mon Sep 17 00:00:00 2001
From: Fabian Vogt <fabian@ritter-vogt.de>
Date: Sat, 15 May 2021 17:45:54 +0200
Subject: [PATCH] Fix informing the underlying widget when leaving
 SplitterProxy

While the SplitterProxy is active, it intercepts all relevant events, so that
the underlying widget still thinks it's in the same "on splitter" state. When
the SplitterProxy is left, the underlying widget is sent a HoverLeave/HoverMove
event to make it aware of the new current cursor position. Without this, it
doesn't know that it's not supposed to be in the "on splitter" state, and when
it regains focus it just re-activates the SplitterProxy at the current cursor
position.

This was broken by accident in d201a1f187 ("Fix SplitterProxy not clearing
when above another QSplitterHandle"), which moved the hide() call past the
call to QCoreApplication::sendEvent. Previously that made isVisible() false,
which also prevented the interception of the HoverLeave/HoverMove events.

BUG: 436473
---
 kstyle/breezesplitterproxy.cpp | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/kstyle/breezesplitterproxy.cpp b/kstyle/breezesplitterproxy.cpp
index 0cf5685f..d4db407b 100644
--- a/kstyle/breezesplitterproxy.cpp
+++ b/kstyle/breezesplitterproxy.cpp
@@ -341,11 +341,14 @@ namespace Breeze
         // send hover event
         if( _splitter )
         {
-            QHoverEvent hoverEvent(
-                qobject_cast<QSplitterHandle*>(_splitter.data()) ? QEvent::HoverLeave : QEvent::HoverMove,
-                _splitter.data()->mapFromGlobal(QCursor::pos()), _hook);
-            QCoreApplication::sendEvent( _splitter.data(), &hoverEvent );
+            // SplitterProxy intercepts HoverLeave/HoverMove events to _splitter,
+            // but this is meant to reach it directly. Unset _splitter to stop interception.
+            auto splitter = _splitter;
             _splitter.clear();
+            QHoverEvent hoverEvent(
+                qobject_cast<QSplitterHandle*>(splitter.data()) ? QEvent::HoverLeave : QEvent::HoverMove,
+                splitter.data()->mapFromGlobal(QCursor::pos()), _hook);
+            QCoreApplication::sendEvent( splitter.data(), &hoverEvent );
         }
 
         // kill timer if any
-- 
GitLab