Open gadfort opened 1 year ago
can't reproduce on macos or linux with the same commit or the ORD head
@jjcherry56 I'll see if I'm able to narrow it down. I just updated the current head and am still getting it.
I get it as well.
It happens when compiled debug. When run under gdb the problem goes away which suggests uninit memory. Valgrind shows
--------------------------------------------------------------------------
==255197== Invalid read of size 2
==255197== at 0x3D26F4C: sta::Vertex::objectIdx() const (Graph.hh:335)
==255197== by 0x3D277CF: sta::ObjectTable<sta::Vertex>::objectId(sta::Vertex const*) (ObjectTable.hh:157)
==255197== by 0x3D23515: sta::Graph::id(sta::Vertex const*) const (Graph.cc:363)
==255197== by 0x3D26E79: sta::VertexIdLess::operator()(sta::Vertex const*, sta::Vertex const*) const (Graph.cc:1611)
==255197== by 0x3BB3A40: std::_Rb_tree<sta::Vertex*, sta::Vertex*, std::_Identity<sta::Vertex*>, sta::VertexIdLess, std::allocator<sta::Vertex*> >::_M_get_insert_unique_pos(sta::Vertex* const&) (stl_tree.h:2054)
==255197== by 0x3E90AA4: std::pair<std::_Rb_tree_iterator<sta::Vertex*>, bool> std::_Rb_tree<sta::Vertex*, sta::Vertex*, std::_Identity<sta::Vertex*>, sta::VertexIdLess, std::allocator<sta::Vertex*> >::_M_insert_unique<sta::Vertex*>(sta::Vertex*&&) (stl_tree.h:2106)
==255197== by 0x3E8F071: std::set<sta::Vertex*, sta::VertexIdLess, std::allocator<sta::Vertex*> >::insert(sta::Vertex*&&) (stl_set.h:520)
==255197== by 0x3E85211: sta::Search::findInputDrvrVertices(sta::VertexSet&) (Search.cc:1620)
==255197== by 0x3E83E56: sta::Search::seedArrivals() (Search.cc:1366)
==255197== by 0x3E82406: sta::Search::findArrivals1() (Search.cc:1002)
==255197== by 0x3E822AC: sta::Search::findArrivals(int, sta::VertexVisitor*) (Search.cc:982)
==255197== by 0x3E820D4: sta::Search::findAllArrivals(sta::VertexVisitor*) (Search.cc:940)
==255197== Address 0x24 is not stack'd, malloc'd or (recently) free'd
==255197==
Signal 11 received
@jjcherry56 is that sufficient for you to take over?
Describe the bug
When I added
check_setup
to my flow before reading the floorplan DEF,report_checks
causes a segfault, while if I don't include thecheck_setup
, it does not segfaultExpected Behavior
No segfault
Environment
To Reproduce
floorplan_gcd_sky130hd_base_2023-01-07_16-21.tar.gz ./run-me-gcd-sky130hd-base.sh
Relevant log output
Screenshots
No response
Additional Context
No response