./xgcc -B./ -B/home/dave/opt/gnu/gcc/gcc-4.1.0/hppa-linux/bin/ -isystem /home/da ve/opt/gnu/gcc/gcc-4.1.0/hppa-linux/include -isystem /home/dave/opt/gnu/gcc/gcc- 4.1.0/hppa-linux/sys-include -L/home/dave/gnu/gcc-4.0/objdir/gcc/../ld -O2 -O2 -g -O2 -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-proto types -Wold-style-definition -isystem ./include -fPIC -DELF=1 -DLINUX=1 -g -DH AVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -I. -I. -I../../gcc/gcc -I../../gcc/gcc/. -I../../gcc/gcc/../include -I../../gcc/gcc/../libcpp/include -DL_gcov_merge_single -c ../../gcc/gcc/libgcov.c -o libgcc/./_gcov_merge_single .o ../../gcc/gcc/libgcov.c: In function '__gcov_merge_single': ../../gcc/gcc/libgcov.c:652: internal compiler error: in default_secondary_reloa d, at targhooks.c:529 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. make[3]: *** [libgcc/./_gcov_merge_single.o] Error 1
I see that you are having a bad couple of weeks.
Subject: Re: [4.2 Regression] libgcov.c:652: ICE: in default_secondary_reload, at targhooks.c:529 > I see that you are having a bad couple of weeks. Yah. Definitely, slows progress on PA specific stuff. Dave
Subject: Re: [4.2 Regression] libgcov.c:652: ICE: in default_secondary_reload, at targhooks.c:529 > I see that you are having a bad couple of weeks. Thing were somewhat better before r107468. Dave
The preprocessed source is where?
Subject: Re: [4.2 Regression] libgcov.c:652: ICE: in default_secondary_reload, at targhooks.c:529 > The preprocessed source is where? Attached. Dave
Created attachment 10349 [details] _gcov_merge_single.i
Subject: Re: [4.2 Regression] libgcov.c:652: ICE: in default_secondary_reload, at targhooks.c:529 This is what I see in gdb: Assembling functions: __gcov_merge_single Breakpoint 3, default_secondary_reload (in_p=0 '\0', x=0x4000d720, reload_class=FP_REGS, reload_mode=DImode, sri=0xbff5ed10) at ../../gcc/gcc/targhooks.c:534 534 gcc_assert (scratch_class == class); (gdb) p class $1 = GENERAL_REGS (gdb) p scratch_class $2 = GENERAL_REGS (gdb) p insn_data[(int) icode] $3 = {name = 0x864b60 "reload_outdi", output = {single = 0x0, multi = 0x0, function = 0}, genfun = @0x8dfda2: 0x8e117c <fixup_branch_offset+15508>, operand = 0x862bc4, n_operands = 3 '\003', n_dups = 0 '\0', n_alternatives = 1 '\001', output_format = 0 '\0'} (gdb) p debug_rtx (x) (mem:DI (reg/f:SI 30 %r30) [0 S8 A64]) $4 = void (gdb) p reload_class $5 = FP_REGS (gdb) p reload_mode $6 = DImode The problem is that the class for the scratch register specified in "reload_outdi" is GENERAL_REGS but we actually need R1_REGS for certain reloads. The documentation indicates that reload_in* and reload_out* have been obsoleted by the new target hook. If that's really the case, then it would appear emit_move_sequence needs a rewrite to somehow integrate it with the new target hook. Dave
Confirmed.
See http://gcc.gnu.org/ml/gcc-patches/2005-11/msg02110.html.