sw UNO API: fix infinite loop when building portions of a bookmark range

The problem was that in case the bookmark range points to a number of
characters in a text node, where neither the start or the end is the paragraph
start and end, and in case there is an at-char anchored object at the beginning
of the paragraph, then lcl_CreatePortions() never ended.

It is assumed that the loop in lcl_CreatePortions() will end at some stage, as
every iteration moves the cursor forward. But this wasn't true in the above
situation: the first frame was anchored at char pos 0, the range was char
positions 1..7, and we failed to ignore frames which are anchored before start:
so position was constantly 1.

Fix the problem by explicitly ignoring frames before the current position, so
lcl_ExportFrames() properly returns -1 when there are no anchored objects in
the current range. Testcase is in the next commit.

Change-Id: I73662e09cd09ee7e0ea4575b1150beb40f1ccc3a

1 files changed, 5 insertions, 0 deletions

diff --git a/sw/source/core/unocore/unoportenum.cxx b/sw/source/core/unocore/unoportenum.cxx
index 039e60f94436..422bce910256 100644
--- a/sw/source/core/unocore/unoportenum.cxx
+++ b/sw/source/core/unocore/unoportenum.cxx
@@ -1193,6 +1193,11 @@ static sal_Int32 lcl_ExportFrames(
     FrameDependSortList_t & i_rFrames,
     sal_Int32 const i_nCurrentIndex)
 {
+    // Ignore frames which are not exported, as we are exporting a selection
+    // and they are anchored before the start of the selection.
+    while (i_rFrames.size() && i_rFrames.front().nIndex < i_nCurrentIndex)
+        i_rFrames.pop_front();
+
     // find first Frame in (sorted) i_rFrames at current position
     while (i_rFrames.size() && (i_rFrames.front().nIndex == i_nCurrentIndex))
     // do not check for i_nEnd here; this is done implicity by lcl_MoveCursor