sw: fix mysterious layout loop in CppunitTest_sw_uiwriter3 testTdf104649

This didn't happen with master from a week ago on Fedora 35, but happens
with Monday's master on Fedora 36. Also happens with libreoffice-7-3
branch.

Fundamentally the problem with the bugdoc is that there are tables in
footnotes, which aren't really supported and can't split across pages
like they would need to.

The loop happens because a footnote on page 48 invalidates position of
its anchor frame 549 on page 45.

This is probably pointless, let's only invalidate if the anchor is on
the same page (it should be on the same page, but probably the tables in
other footnotes get in the way).

Change-Id: I87976c7f8b35725bc8e642133bebb396d37ff0be
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/134877
Tested-by: Jenkins
Reviewed-by: Michael Stahl <michael.stahl@allotropia.de>
(cherry picked from commit 0496252c2c7fd2d694c4a73f387ef75e0021de3e)
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/134894
Reviewed-by: Caolán McNamara <caolanm@redhat.com>

1 files changed, 10 insertions, 1 deletions

diff --git a/sw/source/core/layout/wsfrm.cxx b/sw/source/core/layout/wsfrm.cxx
index 62c99a5f6e87..43bd9a7037a1 100644
--- a/sw/source/core/layout/wsfrm.cxx
+++ b/sw/source/core/layout/wsfrm.cxx
@@ -3001,7 +3001,16 @@ SwTwips SwLayoutFrame::ShrinkFrame( SwTwips nDist, bool bTst, bool bInfo )
                 pTmp->InvalidateSize();
             }
             else
-                pCnt->InvalidatePos();
+            {
+                if (pCnt->FindPageFrame() == FindPageFrame())
+                {
+                    pCnt->InvalidatePos();
+                }
+                else
+                {
+                    SAL_WARN("sw.layout", "footnote frame on different page than ref frame?");
+                }
+            }
         }
     }
     return nReal;