[Bug rtl-optimization/52175] New: [4.7 regression] ICE in maybe_record_trace_start after invalid dbr_schedule transformation

["rsandifo at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52175

             Bug #: 52175
           Summary: [4.7 regression] ICE in maybe_record_trace_start after
                    invalid dbr_schedule transformation
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: rsandifo@gcc.gnu.org


Created attachment 26618
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26618
Testcase

The attached testcase fails on MIPS targets (e.g. mipsisa64-elf
and mips64-linux-gnu) when compiled with -O2 -g:

foo.c: In function âfooâ:
foo.c:25:1: internal compiler error: in maybe_record_trace_start, at
dwarf2cfi.c:2231
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

The problem is that dbr_schedule is applying its PLUS/MINUS
optimisation to a stack adjustment: that is, it fills a
delay slot with a stack adjustment that is only valid on
one edge, then inserts the opposite adjustment on the
other edge.

This is related to PR51471.  We shouldn't do this
adjustment for frame-related instructions.