ggc_mark_rtx and gcc_mark_tree

ggc_mark_rtx and gcc_mark_tree

[Josef Zlomek]

Hi!

I'm updating my 8-month old patch and I have found out
that ggc_mark_rtx_children_1 was removed.
In my old patch I was marking children of my NOTE_INSN,
is it now superfluous to mark them now or
is it necessary to do it some other way?

Thank you for information.

Josef


Here is the chunk that failed patching:

diff -c -3 -p -r1.54 -r1.45.2.11
*** gcc/ggc-common.c    4 Jun 2002 17:32:52 -0000       1.54
--- gcc/ggc-common.c    12 Jun 2002 16:27:26 -0000      1.45.2.11
*************** ggc_mark_rtx_children_1 (r)
***************
*** 222,227 ****
            case NOTE_INSN_BLOCK_BEG:
            case NOTE_INSN_BLOCK_END:
              ggc_mark_tree (NOTE_BLOCK (r));
              break;

            default:
--- 225,236 ----
            case NOTE_INSN_BLOCK_BEG:
            case NOTE_INSN_BLOCK_END:
              ggc_mark_tree (NOTE_BLOCK (r));
+             break;
+             
+           case NOTE_INSN_VAR_LOCATION:
+             ggc_mark_rtx (NOTE_VAR_LOCATION (r));
+             ggc_mark_tree (NOTE_VAR_LOCATION_DECL (r));
+             ggc_mark_rtx (NOTE_VAR_LOCATION_LOC (r));
              break;

            default:

Re: ggc_mark_rtx and gcc_mark_tree

[Geoffrey Keating]

On Wednesday, January 8, 2003, at 04:43  AM, Josef Zlomek wrote:

> Hi!
>
> I'm updating my 8-month old patch and I have found out
> that ggc_mark_rtx_children_1 was removed.
> In my old patch I was marking children of my NOTE_INSN,
> is it now superfluous to mark them now or
> is it necessary to do it some other way?
>
You do need to mark them, but it's probably working already; the new 
marker routines are autogenerated from rtl.def.  gengtype will abort if 
it detects a rtx_format entry that it doesn't understand, so if it 
completes then you don't need to do anything.