On February 7, 2019 11:37:29 PM GMT+01:00, Jakub Jelinek <ja...@redhat.com>
wrote:
>Hi!
>
>I've noticed pre_and_rev_post_order_compute_fn is one of (apparently
>many)
>functions that take a struct function * argument and accept any
>function, as
>long as it is cfun. For a patch I've been working on I actually need
>it
>to handle other functions as well and in this case it is trivial to
>fix.
>I've seen several others e.g. in graph.c, though some of those are
>harder to
>deal with, as some there are calls to functions that don't have a
>variant
>with struct function * argument and assume cfun implicitly.
>
>Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
EH...
OK.
Richard.
>2019-02-07 Jakub Jelinek <ja...@redhat.com>
>
> * cfganal.c (pre_and_rev_post_order_compute_fn): Use fn instead of
> cfun everywhere.
>
>--- gcc/cfganal.c.jj 2019-01-01 12:37:17.915962514 +0100
>+++ gcc/cfganal.c 2019-02-07 20:47:29.035520143 +0100
>@@ -951,10 +951,10 @@ pre_and_rev_post_order_compute_fn (struc
> bool include_entry_exit)
> {
> int pre_order_num = 0;
>- int rev_post_order_num = n_basic_blocks_for_fn (cfun) - 1;
>+ int rev_post_order_num = n_basic_blocks_for_fn (fn) - 1;
>
> /* Allocate stack for back-tracking up CFG. */
>- auto_vec<edge_iterator, 20> stack (n_basic_blocks_for_fn (cfun) +
>1);
>+ auto_vec<edge_iterator, 20> stack (n_basic_blocks_for_fn (fn) + 1);
>
> if (include_entry_exit)
> {
>@@ -968,7 +968,7 @@ pre_and_rev_post_order_compute_fn (struc
> rev_post_order_num -= NUM_FIXED_BLOCKS;
>
> /* Allocate bitmap to track nodes that have been visited. */
>- auto_sbitmap visited (last_basic_block_for_fn (cfun));
>+ auto_sbitmap visited (last_basic_block_for_fn (fn));
>
> /* None of the nodes in the CFG have been visited yet. */
> bitmap_clear (visited);
>
> Jakub
