https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82009
Jerry DeLisle <jvdelisle at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jvdelisle at gcc dot gnu.org
--- Comment #5 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
This seems on the verge of silly. The ICE is at an assert:
assert (saved_local_decls == NULL_TREE);
When the function gfc_process_block_locals concludes it does
saved_local_decls = NULL_TREE;
So as I looked at this I thought it is expected to be already set at NULL_TREE
and it leaves it that way. Why not just set it instead of doing the assert.
So I did this:
diff --git a/gcc/fortran/trans-decl.c b/gcc/fortran/trans-decl.c
index 254768c5828..08c1ebd2d4b 100644
--- a/gcc/fortran/trans-decl.c
+++ b/gcc/fortran/trans-decl.c
@@ -6751,7 +6751,7 @@ gfc_process_block_locals (gfc_namespace* ns)
{
tree decl;
- gcc_assert (saved_local_decls == NULL_TREE);
+ saved_local_decls = NULL_TREE;
has_coarray_vars = false;
generate_local_vars (ns);
The problem goes away and it regression tests fine.
saved_local_decls is used only in one other place where it is set in
add_decl_as_local(). It is not reset anywhere else. So I think the pacth above
is likely correct.
