------- Comment #3 from amylaar at gcc dot gnu dot org 2007-05-17 17:50 -------
(In reply to comment #2)
> Do you still have the testcase?
>
Not as such. However, AFAICR it was triggered by a regression test
(i.e. building library or testcase).
But the issue should be clear enough without a testcase; it's explained in a
comment of the patch.
While the current code of noce_emit_move_insn has a check for recog_memoized,
It will still emit en invalid move if the move is invalid because the
destination is a paradoxical SUBREG with TRULY_NOOP_TRUNCATION being false,
and source is a REG.
--
amylaar at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |NEW
Ever Confirmed|0 |1
Last reconfirmed|0000-00-00 00:00:00 |2007-05-17 17:50:53
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20369
