https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101047
--- Comment #5 from Paul Thomas <pault at gcc dot gnu.org> --- Created attachment 61121 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=61121&action=edit An update of Jose's patch The attached is an updated version of Jose's that applies to the current trunk. I am either missing another patch or this one is not complete. Not only does the original testcase fail on associated with both the 'pointer' and the 'target' arguments (the 'dim' fields of 'pointer' are not set) but regression testing reveals: FAIL: gfortran.dg/class_allocate_14.f90 -O0 scan-tree-dump-times original "static struct __class_m_T_1_0a b = {._data={.data=0B, .dtype={.elem_len=4, .rank=1, .type=5}}, ._vptr=&__vtab_m_T};" 1 [The dump shows static struct __class_m_T_1_0a b = {._data={.data=0B, .dtype={.elem_len=4, .version=0, .rank=1, .type=5}}, ._vptr=&__vtab_m_T}; ie. .version=0, must be added] ...snip... FAIL: gfortran.dg/init_workout.f90 -O0 execution test [of course! => stop 3] ...snip... FAIL: gfortran.dg/size_dim_2.f90 -O (test for excess errors) [lines 16 & 18, which flang-new and ifx also pick up] Paul
