diff options
Diffstat (limited to 'gcc/gimple.c')
| -rw-r--r-- | gcc/gimple.c | 11 |
1 files changed, 10 insertions, 1 deletions
diff --git a/gcc/gimple.c b/gcc/gimple.c index 1ff9b3a..e5dc184 100644 --- a/gcc/gimple.c +++ b/gcc/gimple.c @@ -3481,11 +3481,20 @@ gimple_types_compatible_p (tree t1, tree t2) && RECORD_OR_UNION_TYPE_P (TREE_TYPE (t1)) && (!COMPLETE_TYPE_P (TREE_TYPE (t1)) || !COMPLETE_TYPE_P (TREE_TYPE (t2))) + && TYPE_QUALS (TREE_TYPE (t1)) == TYPE_QUALS (TREE_TYPE (t2)) && compare_type_names_p (TYPE_MAIN_VARIANT (TREE_TYPE (t1)), TYPE_MAIN_VARIANT (TREE_TYPE (t2)), true)) { /* Replace the pointed-to incomplete type with the - complete one. */ + complete one. + ??? This simple name-based merging causes at least some + of the ICEs in canonicalizing FIELD_DECLs during stmt + read. For example in GCC we have two different struct deps + and we mismatch the use in struct cpp_reader in sched-int.h + vs. mkdeps.c. Of course the whole exercise is for TBAA + with structs which contain pointers to incomplete types + in one unit and to complete ones in another. So we + probably should merge these types only with more context. */ if (COMPLETE_TYPE_P (TREE_TYPE (t2))) TREE_TYPE (t1) = TREE_TYPE (t2); else |