The failure message is: stage1/xgcc -Bstage1/ -B/opt/build/eric/local/gcc/sparc-sun-solaris2.10/bin/ -c -g -O2 -fprofile-use -freorder-blocks-and-partition -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long -Wno-variadic-macros -Wold-style-definition -Wmissing-format-attribute -Werror -fno-common -DHAVE_CONFIG_H -I. -I. -I/home/eric/cvs/gcc/gcc -I/home/eric/cvs/gcc/gcc/. -I/home/eric/cvs/gcc/gcc/../include -I./../intl -I/home/eric/cvs/gcc/gcc/../libcpp/include -I/opt/build/eric/local/include -I/opt/build/eric/local/include /home/eric/cvs/gcc/gcc/attribs.c -o attribs.o /home/eric/cvs/gcc/gcc/attribs.c:332: internal compiler error: Segmentation Fault Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. gmake[2]: *** [attribs.o] Error 1 gmake[2]: Leaving directory `/opt/build/eric/gcc/gcc' gmake[1]: *** [stagefeedback_build] Error 2 gmake[1]: Leaving directory `/opt/build/eric/gcc/gcc' gmake: *** [profiledbootstrap] Error 2 A bit annoying to debug as the stage1 compiler is apparently built without debug info. The bug happens with -g -fprofile-use -freorder-blocks-and-partition.
Stage1 should be built at -O0 -g.
> Stage1 should be built at -O0 -g. It's probably not the regular stage1. I'm bootstrapping again to watch what is going on exactly.
Hmm: stageprofile_build: stage1_copy It is the normal stage1. If you cannot figure out the backtrace, I might be able to reproduce this by a cross compiler. Could you attach the .i file, .gcda and .gcno files?
> It is the normal stage1. No, it isn't, look at the options. Normal stage1 has -g -O2 -fprofile-generate. It's very likely stage1 of stagefeedback_build. I'm replaying the bootstrap.
Here's the nice version: stage1/xgcc -Bstage1/ -B/opt/build/eric/local/gcc/sparc-sun-solaris2.10/bin/ -c -g -O2 -fprofile-use -freorder-blocks-and-partition -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long -Wno-variadic-macros -Wold-style-definition -Wmissing-format-attribute -Werror -fno-common -DHAVE_CONFIG_H -I. -I. -I/home/eric/cvs/gcc/gcc -I/home/eric/cvs/gcc/gcc/. -I/home/eric/cvs/gcc/gcc/../include -I./../intl -I/home/eric/cvs/gcc/gcc/../libcpp/include -I/opt/build/eric/local/include -I/opt/build/eric/local/include /home/eric/cvs/gcc/gcc/attribs.c -o attribs.o /home/eric/cvs/gcc/gcc/attribs.c: In function 'decl_attributes': /home/eric/cvs/gcc/gcc/attribs.c:332: internal compiler error: RTL check: expected elt 3 type 'B', have '0' (rtx barrier) in fix_crossing_unconditional_branches, at bb-reorder.c:1742 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. gmake[2]: *** [attribs.o] Error 1
Investigating.
Subject: Bug 24460 CVSROOT: /cvs/gcc Module name: gcc Changes by: ebotcazou@gcc.gnu.org 2005-10-21 21:44:42 Modified files: gcc : ChangeLog bb-reorder.c Log message: PR rtl-optimization/24460 * bb-reorder.c (fix_crossing_unconditional_branches): Do not set the basic block for barriers. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.10202&r2=2.10203 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/bb-reorder.c.diff?cvsroot=gcc&r1=1.114&r2=1.115
Now the failure is: Program received signal SIGSEGV, Segmentation fault. 0x000000000067d51c in output_line_info () at /home/eric/cvs/gcc/gcc/dwarf2out.c:7902 7902 if (function != line_info->function) (gdb) bt #0 0x000000000067d51c in output_line_info () at /home/eric/cvs/gcc/gcc/dwarf2out.c:7902 #1 0x000000000068e101 in dwarf2out_finish ( filename=0x7fbffff281 "../attribs.i") at /home/eric/cvs/gcc/gcc/dwarf2out.c:14082 #2 0x00000000009ecdb5 in compile_file () at /home/eric/cvs/gcc/gcc/toplev.c:1028 #3 0x00000000009ee58d in do_compile () at /home/eric/cvs/gcc/gcc/toplev.c:1945 #4 0x00000000009ee5ef in toplev_main (argc=5, argv=0x7fbfffee48) at /home/eric/cvs/gcc/gcc/toplev.c:1977 #5 0x000000000049ae8b in main (argc=5, argv=0x7fbfffee48) at /home/eric/cvs/gcc/gcc/main.c:35
Subject: Bug 24460 CVSROOT: /cvs/gcc Module name: gcc Changes by: ebotcazou@gcc.gnu.org 2005-10-26 07:03:32 Modified files: gcc : ChangeLog bb-reorder.c dwarf2out.c output.h varasm.c Log message: PR rtl-optimization/24460 * dwarf2out.c (have_switched_text_sections): New boolean variable. (dwarf2out_switch_text_section): Set it to true instead of incrementing separate_line_info_table_in_use. (output_loc_list): Additionally test have_switched_text_sections. (output_ranges): Likewise. (dwarf2out_finish): Likewise. * varasm.c (assemble_start_function): Do not call insert_section_boundary_note. (assemble_end_function): If flag_reorder_blocks_and_partition, switch to the function's section before emitting the .size directive. * bb-reorder.c (insert_section_boundary_note): Staticify. (rest_of_handle_reorder_blocks): Call insert_section_boundary_note. * output.h (insert_section_boundary_note): Delete. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.10214&r2=2.10215 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/bb-reorder.c.diff?cvsroot=gcc&r1=1.115&r2=1.116 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/dwarf2out.c.diff?cvsroot=gcc&r1=1.615&r2=1.616 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/output.h.diff?cvsroot=gcc&r1=1.161&r2=1.162 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/varasm.c.diff?cvsroot=gcc&r1=1.534&r2=1.535
See http://gcc.gnu.org/ml/gcc-patches/2005-10/msg01409.html