I have set ACCEPT_KEYWORDS="~x86" module-init-tools and modutils build fine after I do hardened-gcc -R but fail to build after I do hardened-gcc -A . . gcc -O3 -march=athlon-xp -pipe -fstack-protector -I./../include -D_GNU_SOURCE -DCONFIG_ROOT_CHECK_OFF=0 -DELF_MACHINE_H='"elf_i386.h"' -DARCH_i386 -c -o sys_oim.o sys_oim.c sys_oim.c: In function `old_sys_init_module': sys_oim.c:38: can't find a register in class `BREG' while reloading `asm' make[1]: *** [sys_oim.o] Error 1 make[1]: *** Waiting for unfinished jobs.... make[1]: Leaving directory `/var/tmp/portage/module-init-tools-0.9.13_pre2/work/modutils-2.4.25/util' make: *** [all] Error 2 It seems to be the same with earlier versions of module-init-tools. I've had a play with the code and found that if I left sys_oim.o out of the library, nothing seemed to miss it. Reproducible: Always Steps to Reproduce: 1.emerge hardened-gcc 2.hardened-gcc -A 3.emerge module-init-tools Actual Results: failed to compile, as described under 'details' Expected Results: successful compilation System uname: 2.6.0-test4 i686 AMD Athlon(tm) XP 1800+ ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-O3 -march=athlon-xp -pipe -fstack-protector" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/kde/2/share/config /usr/kde/3/share/config /usr/X11R6/lib/X11/xkb /usr/kde/3.1/share/config /usr/share/config" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-O3 -march=athlon-xp -pipe -fstack-protector" DISTDIR="/usr/portage/distfiles" FEATURES="sandbox ccache autoaddcvs" GENTOO_MIRRORS="ftp://cs.ubishops.ca/pub/gentoo ftp://ftp.ucsb.edu/pub/mirrors/linux/gentoo/ ftp://mirror.iawnet.sandia.gov/pub/gentoo/" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="" SYNC="rsync://rsync.europe.gentoo.org/gentoo-portage" USE="x86 oss apm encode gnome libg++ mikmod xmms gdbm berkdb svga tcltk sdl libwww esd imlib oggvorbis gtk 3dnow acpi alsa arts avi cdr crypt cups dga directfb dvd foomaticdb gif gpm java jpeg kde kerberos ldap mbox mad mmx motif mozilla mpeg mpi ncurses nls opengl pam pdflib perl png python qt quicktime readline sasl slang spell sse ssl tcpd tetex tiff truetype usb wmf X xml xml2 xv zlib"
Please bump to the hardened team to fix this.
I think this is somewhat of a common problem for code that uses inline asm for system calls (util-linux had the suffered the problem before) Note: The hardened-gcc has alot of logic built into that enables transparent -fstack-protector. So you should be able to remove -fstack-protector assigning bug to bug pappy@gentoo.org pappy can you look into this and make module-init-tools and modutils -fPIC aware.
asm inline constructed syscalls (found in util-linux also) are prohibited by the transparent -fPIC in the specs file of hardened-gcc same problem applies to util-linux where a PIC patch has been designed by solar and pappy please do not CC azarah as he is a busy man and hardened-gcc is not his toy but mine.
There seems to be similar problems compiling libmpeg3 gcc -c `cat i686/c_flags` video/slice.c -o i686/video/slice.o /usr/include/stdlib.h: In function `mpeg3video_addblock': video/slice.c:211: can't find a register in class `BREG' while reloading `asm' make: *** [i686/video/slice.o] Error 1 !!! ERROR: media-libs/libmpeg3-1.5.1 failed. !!! Function src_compile, Line 40, Exitcode 2 !!! (no error message) and mplayer gcc -O4 -march=athlon-xp -mcpu=athlon-xp -pipe -ffast-math -fomit-frame-pointer -D_REENTRANT -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -g -DHAVE_AV_CONFIG_H -I.. -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE -c -o msmpeg4.o msmpeg4.c In file included from msmpeg4.c:2007: wmv2.c: In function `wmv2_encode_end': wmv2.c:98: warning: unused variable `w' In file included from msmpeg4.c:2007: wmv2.c: In function `wmv2_add_block': wmv2.c:578: warning: unused variable `temp' wmv2.c:579: warning: unused variable `i' wmv2.c: In function `msmpeg4_pred_dc': msmpeg4.c:703: can't find a register in class `BREG' while reloading `asm' make[1]: *** [msmpeg4.o] Error 1 make[1]: Leaving directory `/var/tmp/portage/mplayer-0.91/work/MPlayer-0.91/libavcodec' make: *** [libavcodec/libavcodec.a] Error 2 !!! ERROR: media-video/mplayer-0.91 failed. !!! Function src_compile, Line 285, Exitcode 2 !!! Failed to build MPlayer!
This is still failing... modutils is apart of the "system" in just about every profile in gentoo so I think a workaround needs to be put in place rather quickly. pappy any luck with some workaround logic?
made the changes to the ebuilds, please test sunday night and report back, thanks, Alex
Laurence, could you close the bug after you verified the packages? TIA, Alex
Sorry for the delay. Unfortunately I can't verify the packages as I still can't build module-init-tools, but now for a different reason. emerge sync and emerge -Du world have upgraded gcc to 3.3.1, which may have caused the new problem. It is also somewhat odd in that the build does not always fail at the same point. Below shows it failing at lsmod, but sometimes it is insmod or modinfo, which is confusing me. Shall I close this bug anyway and do I need to file a new bug or is this new thing something silly I've done? gcc -mcpu=athlon-xp -O3 -pipe -yet_exec -Wunused -Wall -Wcast-align -o lsmod lsmod.o /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0xd9): In function `__stack_smash_handler': /var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2050: undefined refer ence to `sigfillset' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0xe9):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2051: undefine d reference to `sigdelset' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x12b):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2058: undefin ed reference to `strncat' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x1b7):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2066: undefin ed reference to `socket' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x1f2):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2072: undefin ed reference to `strncpy' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x226):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2075: undefin ed reference to `sendto' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x252):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2085: undefin ed reference to `sigfillset' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x2ba):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2092: undefin ed reference to `strncat' /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.1/libgcc.a(_stack_smash_handler.oS)(.text +0x327):/var/tmp/portage/gcc-3.3.1-r1/work/gcc-3.3.1/gcc/libgcc2.c:2061: undefin ed reference to `strncat' collect2: ld returned 1 exit status make: *** [insmod.static] Error 1 make: *** Waiting for unfinished jobs.... !!! ERROR: sys-apps/module-init-tools-0.9.13_pre2 failed. !!! Function src_compile, Line 87, Exitcode 2 !!! emake module-init-tools failed
another one got hit by the gcc upgrade? try to copy the `gcc-config -L`/libgcc*.so into /lib and run ldconfig
Doesn't seem to help, I'm afraid. I did suffer the problems mentioned in bug 28185, but had managed to sort that out. The current problem seems to be different. If it's only me that has this problem then I shall assume that I've done something odd to my system and will sort this out. I guess I should close this bug as the -yet_exec flag seems now to be present and I imagine this will solve the original problem.
I don't think I'm capable of closing the bug. Am I supposed to be able to do this? The 'Status' and 'Resolution' fields do not seem to allow editing. Sorry.
Laurence, No only developers can close bugs. Alex (pappy) has to change the status to TEST-REQUEST,CLOSED or reassign the bug when he is ready.
i only have "Leave as new" and cannot "change the bug" to FIXED someone ought to look at that i guess
changing resolution to FIXED
