[Bug middle-end/89150] [9 regression] Tree form bitmaps break GC

[nathan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89150

--- Comment #3 from Nathan Sidwell <nathan at gcc dot gnu.org> ---
I didn't know there were no tree-form bitmaps yet.

Contrary to my original assertion, I think just dropping the chain_prev
("%h.prev") marker will suffice.  Normal use of a list-form bitmap starts at
the head.  The only way to keep hold of an internal pointer is by looking
behind the API.

For tree-form bitmaps we'd iterate down the right-hand branch and recurse on
the left hand ones.