https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80850
--- Comment #17 from Paul Thomas <pault at gcc dot gnu.org> --- (In reply to Paul Thomas from comment #16) > (In reply to Jerry DeLisle from comment #15) > > (In reply to Paul Thomas from comment #14) > > > PS If I remove the STAT= from the allocate, the code runs just fine in DEV > > > mode. > > > > > > Paul > > > > I have not looke at this one. Have you tried with valgrind or sanitize tools > > to see if there are any further hints? > > You will see from the above that the reporter himself has done this already. > Just for the record, the backtrace from the error is: > > Program aborted. Backtrace: > #0 0x4595c4 in __stsubs_MOD_clone_object > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > stsubs.F90:305 > #1 0x470f0e in __gfc_base_MOD_contelemconstruct > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > gfc_base.F90:354 > #2 0x438586 in __gfc_vector_MOD_vectorelemconstruct > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > gfc_vector.F90:211 > #3 0x4266cd in __gfc_vector_MOD_vectoriterappend > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > gfc_vector.F90:807 > #4 0x44a83e in __gfc_graph_MOD_graphiterappendvertex > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > gfc_graph.F90:1113 > #5 0x45022e in __gfc_graph_test_MOD_test_gfc_graph > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > gfc_graph.F90:1402 > #6 0x403ac1 in MAIN__ > at > /home/pault/prs/pr80850/GFC-master-790d92e9ccc22c559db4118f2b8a51170c135b19/ > main.F90:94 > #7 0x403ed5 in main > > gfc_vector.F90:807 contains a call that looks horribly like the problem line > in PR81758 > call > this%container%vec_tile(q(4))%tile_batch(q(3))%batch_seg(q(2))%seg_elem(q(1)) > %& > &vector_elem_ctor(elem_val,ierr,assoc_only=assoc) > > However, it is definitely the allocate that is the problem as evidenced by > it disappearing if STAT is not used. > > Unfortunately, as Dmitry reports above, the fault also goes away in gdb :-( > > Cheers > > Paul I instrumented trans.c:(gfc_allocate_using_malloc) by setting the STAT variable to the requested size, when an allocation error is detected. When the code fails, this is a typical outcome: >>>>> errc = 1713129088 #ERROR(stsubs:clone_object): sourced allocate() failed: Attempt to allocate an allocated object Somewhere, use is being made of an uninitialized variable upstream from stsubs.F90(clone_object). Paul
