[Bug target/62254] [4.9/5/6 Regression] gcc-4.9 ICEs on linux kernel zlib for armv3

[arnd at linaro dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62254

Arnd Bergmann <arnd at linaro dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |arnd at linaro dot org

--- Comment #9 from Arnd Bergmann <arnd at linaro dot org> ---
Created attachment 38116
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38116&action=edit
preprocessed source for another test, not minimized

I'm still getting what looks like the same or very similar problem with
arm-linux-gnueabi-gcc (GCC) 6.0.0 20160323:

arm-linux-gnueabi-gcc -march=armv3 -Os -fno-omit-frame-pointer
-fstack-protector-strong -Wno-pointer-sign -fprofile-arcs  -fno-tree-loop-im 
-c comm.i


/git/arm-soc/drivers/block/paride/comm.c: In function 'comm_write_block':
/git/arm-soc/drivers/block/paride/comm.c:177:1: internal compiler error: in
arm_reload_out_hi, at config/arm/arm.c:15604
 }
 ^
0xbba4e2 arm_reload_out_hi(rtx_def**)
        /home/arnd/git/gcc/gcc/config/arm/arm.c:15604
0xcda654 gen_reload_outhi(rtx_def*, rtx_def*, rtx_def*)
        /home/arnd/git/gcc/gcc/config/arm/arm.md:6408
0x831c71 insn_gen_fn::operator()(rtx_def*, rtx_def*, rtx_def*) const
        /home/arnd/git/gcc/gcc/recog.h:302
0x831c71 check_and_process_move
        /home/arnd/git/gcc/gcc/lra-constraints.c:1217
0x831c71 curr_insn_transform
        /home/arnd/git/gcc/gcc/lra-constraints.c:3512
0x8328de lra_constraints(bool)
        /home/arnd/git/gcc/gcc/lra-constraints.c:4494
0x8234e4 lra(_IO_FILE*)
        /home/arnd/git/gcc/gcc/lra.c:2290
0x7e3ec1 do_reload
        /home/arnd/git/gcc/gcc/ira.c:5425
0x7e3ec1 execute
        /home/arnd/git/gcc/gcc/ira.c:5609
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.

I've see similar ICEs in around two dozen files in the kernel now, but haven't
checked  which ones remain broken with current gcc-6. Please see if you can
spot something in the test case above, or let me know if you need me to check
the other files or reduce the test case further. Note that the command line
arguments above (-march=armv3 -Os -fno-omit-frame-pointer
-fstack-protector-strong -fprofile-arcs  -fno-tree-loop-im) all seem to be
significant here.