> On Tue, Oct 01, 2019 at 07:04:53PM +0200, Jan Hubicka wrote:
> > * gcc.dg/tree-ssa/ssa-thread-12.c: Mark compure_idf inline.
>
> > --- testsuite/gcc.dg/tree-ssa/ssa-thread-12.c (revision 276272)
> > +++ testsuite/gcc.dg/tree-ssa/ssa-thread-12.c (working copy)
> > @@ -56,7 +56,7 @@ bmp_iter_and_compl (bitmap_iterator * bi
> > }
> >
> > extern int VEC_int_base_length (VEC_int_base *);
> > -bitmap
> > +inline bitmap
> > compute_idf (bitmap def_blocks, bitmap_head * dfs)
> > {
> > bitmap_iterator bi;
>
> Neither this change nor the followup in http://gcc.gnu.org/r276418
> actually works.
> PASS: gcc.dg/tree-ssa/ssa-thread-12.c (test for excess errors)
> gcc.dg/tree-ssa/ssa-thread-12.c: dump file does not exist
> UNRESOLVED: gcc.dg/tree-ssa/ssa-thread-12.c scan-tree-dump thread2 "FSM"
> gcc.dg/tree-ssa/ssa-thread-12.c: dump file does not exist
> UNRESOLVED: gcc.dg/tree-ssa/ssa-thread-12.c scan-tree-dump thread3 "FSM"
> gcc.dg/tree-ssa/ssa-thread-12.c: dump file does not exist
> UNRESOLVED: gcc.dg/tree-ssa/ssa-thread-12.c scan-tree-dump thread4 "FSM"
>
> We shouldn't really have UNRESOLVED testcases in the testsuite.
>
> The reason is that with the static __inline__ bitmap in there, no function
> is actually emitted as nothing uses it, so there are no dumps other than
> original (and 4 empty ones) with -fdump-tree-all.
Aha, sorry for that. I was looking for FAILs and missed the UNRESOLVED.
I will revert that change and go with --param then. The testcase is
somewhat fragile depending on particular chain of inliner decisions.
Honza
>
> Jakub
