diff options
| author | Miklos Vajna <vmiklos@collabora.co.uk> | 2018-02-13 11:44:36 +0100 |
|---|---|---|
| committer | Michael Stahl <mstahl@redhat.com> | 2018-02-16 10:46:38 +0100 |
| commit | ea3111651ee4e88f7abc359ce8f8eb390bbe7d45 (patch) | |
| tree | 11a018ac5895f054b75f346218db985bee6d80cc /sw | |
| parent | 831c4f1874f01bcd7282cdf08b1446acc5c4760e (diff) | |
sw: work around what seems to be a gcc-4.8 compiler bug
Thread 1 "soffice.bin" received signal SIGSEGV, Segmentation fault.
(anonymous namespace)::print_type<15> (ctx=..., info=0x7fffffff7f90, unknown_name=...) at ../../../../../libstdc++-v3/src/c++11/debug.cc:641
641 ../../../../../libstdc++-v3/src/c++11/debug.cc: No such file or directory.
(gdb) up
#1 0x00007ffff6abb008 in (anonymous namespace)::print_description (ctx=..., inst=...) at ../../../../../libstdc++-v3/src/c++11/debug.cc:817
817 in ../../../../../libstdc++-v3/src/c++11/debug.cc
(gdb)
#2 0x00007ffff6abc7aa in (anonymous namespace)::print_description (param=..., ctx=...) at ../../../../../libstdc++-v3/src/c++11/debug.cc:835
835 in ../../../../../libstdc++-v3/src/c++11/debug.cc
(gdb)
#3 __gnu_debug::_Error_formatter::_M_error (this=0x7fffffff7af0) at ../../../../../libstdc++-v3/src/c++11/debug.cc:1061
1061 in ../../../../../libstdc++-v3/src/c++11/debug.cc
(gdb)
#4 0x00007fffca2b6313 in __gnu_debug::operator!=<__gnu_cxx::__normal_iterator<SwAnchoredObject* const*, std::__cxx1998::vector<SwAnchoredObject*, std::allocator<SwAnchoredObject*> > >, std::__debug::vector<SwAnchoredObject*, std::allocator<SwAnchoredObject*> > > (__lhs=0x0, __rhs=invalid iterator) at /usr/include/c++/4.8/debug/safe_iterator.h:535
535 _GLIBCXX_DEBUG_VERIFY(! __lhs._M_singular() && ! __rhs._M_singular(),
(gdb)
#5 0x00007fffca7f3de2 in SwAnchoredObject::UpdateObjInSortedList (this=0x1fa6fd8) at /git/libreoffice/master/sw/source/core/layout/anchoredobject.cxx:629
629 for (SwAnchoredObject* pAnchoredObj : *pObjs)
Given that valgrind doesn't point out anything and the old and the new code is
meant to be the same, my only guess is that the baseline gcc has some codegen
bug. This happens reasonably frequently, current bugdoc was a DOCX file
with 4 shapes anchored to the same paragraph, affecting only dbgutil
builds, it seems.
The tdf#115719 bugdoc is a reproducer for the crash.
(cherry picked from commit d34dcde1d0ccee2d78eea08185f12949f53ceae3)
Change-Id: I2316e25eea87f2aa5736576d5168e113480f80e4
Reviewed-on: https://gerrit.libreoffice.org/49835
Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Michael Stahl <mstahl@redhat.com>
Diffstat (limited to 'sw')
| -rw-r--r-- | sw/source/core/layout/anchoredobject.cxx | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/sw/source/core/layout/anchoredobject.cxx b/sw/source/core/layout/anchoredobject.cxx index 4b79a357771f..dc85d1d235a0 100644 --- a/sw/source/core/layout/anchoredobject.cxx +++ b/sw/source/core/layout/anchoredobject.cxx @@ -625,8 +625,9 @@ void SwAnchoredObject::UpdateObjInSortedList() { const SwSortedObjs* pObjs = GetAnchorFrame()->GetDrawObjs(); // determine start index - for (SwAnchoredObject* pAnchoredObj : *pObjs) + for (auto it = pObjs->begin(); it != pObjs->end(); ++it) { + SwAnchoredObject* pAnchoredObj = *it; if ( pAnchoredObj->ConsiderObjWrapInfluenceOnObjPos() ) pAnchoredObj->InvalidateObjPosForConsiderWrapInfluence(); else |