Add gcc 4.5.1
This is the first pass at adding gcc 4.5.1 to buildroot. It builds fine for the host and the target. The question I have, is the implementation the correct way to change the toolchain? I couldn't find a lot of documentation regarding the toolchain (as far as changing or updating it). Also, can the toolchain mk files be updated to the new format or are their builds too complex?
[PATCH 1/3] dbus: bump to 1.4.0
[PATCH 2/3] m4: bump to 1.4.15
[PATCH 3/3] Added gcc 4.5.1 to buildroot
PATCH 1/3 dbus: bump to 1.4.0
--- .../dbus/dbus-1.2.12-launch-helper-xml-fix.patch | 33 -------------------- package/dbus/dbus.mk | 2 +- 2 files changed, 1 insertions(+), 34 deletions(-) delete mode 100644 package/dbus/dbus-1.2.12-launch-helper-xml-fix.patch
package/dbus/dbus-1.2.12-launch-helper-xml-fix.patch deleted +0 -33
@@ -1,33 +0,0 @@ -dbus-daemon-launch-helper: fix build with libxml - -dbus-daemon-launch-helper links with the trival config parser, which -doesn't provide bus_config_parser_check_doctype(), like the big parser does. -The libxml loader unfortunately wants to call it, so this breaks the build. - -Fix it by including a dummy bus_config_parser_check_doctype() in the trivial -config parser. - -Signed-off-by: Peter Korsgaard <[email protected]> ---- - bus/config-parser-trivial.c | 8 ++++++++ - 1 file changed, 8 insertions(+) - -Index: dbus-1.2.12/bus/config-parser-trivial.c -=================================================================== ---- dbus-1.2.12.orig/bus/config-parser-trivial.c -+++ dbus-1.2.12/bus/config-parser-trivial.c -@@ -310,6 +310,14 @@ - return &parser->service_dirs; - } - -+dbus_bool_t -+bus_config_parser_check_doctype (BusConfigParser *parser, -+ const char *doctype, -+ DBusError *error) -+{ -+ return TRUE; -+} -+ - #ifdef DBUS_BUILD_TESTS - #include <stdio.h> - #include "test.h"
package/dbus/dbus.mk +1 -1
@@ -3,7 +3,7 @@ # dbus # ############################################################# -DBUS_VERSION = 1.2.16 +DBUS_VERSION = 1.4.0 DBUS_SOURCE = dbus-$(DBUS_VERSION).tar.gz DBUS_SITE = http://dbus.freedesktop.org/releases/dbus/ DBUS_LIBTOOL_PATCH = NO
Signature · 1 line
1.7.2.3
PATCH 2/3 m4: bump to 1.4.15
In reply to Martin Banky--- package/m4/m4-1.4.15-00-MB_CUR_MAX.patch | 13 ++++++++++++ package/m4/m4-1.4.15-01-spawn-sched_param.patch | 20 +++++++++++++++++++ package/m4/m4-1.4.8-001-MB_CUR_MAX.patch | 13 ------------ package/m4/m4-1.4.9-002-no-wchar.patch | 24 ----------------------- package/m4/m4.mk | 6 +---- 5 files changed, 34 insertions(+), 42 deletions(-) create mode 100644 package/m4/m4-1.4.15-00-MB_CUR_MAX.patch create mode 100644 package/m4/m4-1.4.15-01-spawn-sched_param.patch delete mode 100644 package/m4/m4-1.4.8-001-MB_CUR_MAX.patch delete mode 100644 package/m4/m4-1.4.9-002-no-wchar.patch
package/m4/m4-1.4.15-00-MB_CUR_MAX.patch new file +13 -0
@@ -0,0 +1,13 @@ +--- m4-1.4.8.orig/lib/regcomp.c 2006-09-24 06:28:37.000000000 +0200 ++++ m4-1.4.8/lib/regcomp.c 2007-01-06 20:53:10.000000000 +0100 +@@ -871,6 +871,10 @@ init_dfa (re_dfa_t *dfa, size_t pat_len) + dfa->state_table = calloc (sizeof (struct re_state_table_entry), table_size); + dfa->state_hash_mask = table_size - 1; + ++#ifndef MB_CUR_MAX ++#define MB_CUR_MAX (0) ++#endif ++ + dfa->mb_cur_max = MB_CUR_MAX; + #ifdef _LIBC + if (dfa->mb_cur_max == 6
package/m4/m4-1.4.15-01-spawn-sched_param.patch new file +20 -0
@@ -0,0 +1,20 @@ +"Apparently, glibc's headers expose the full definition of "struct +sched_param" in cases not required by the standard, and gnulib attempts to +optimize based on this. uClibc does not share glibc's behavior in this one +case, but since it defines __GLIBC__, gnulib sees no need for caution." + +Per Michael Deutschmann + +See http://lists.busybox.net/pipermail/uclibc/2010-March/043721.html + +--- bison-2.4.3-orig/lib/spawn.in.h 2010-04-11 11:31:37.000000000 -0700 ++++ bison-2.4.3/lib/spawn.in.h 2010-09-18 15:42:05.000000000 -0700 +@@ -31,7 +31,7 @@ + + /* Get definitions of 'struct sched_param' and 'sigset_t'. + But avoid namespace pollution on glibc systems. */ +-#ifndef __GLIBC__ ++#if defined __UCLIBC__ || !defined __GLIBC__ + # include <sched.h> + # include <signal.h> + #endif
package/m4/m4-1.4.8-001-MB_CUR_MAX.patch deleted +0 -13
@@ -1,13 +0,0 @@ ---- m4-1.4.8.orig/lib/regcomp.c 2006-09-24 06:28:37.000000000 +0200 -+++ m4-1.4.8/lib/regcomp.c 2007-01-06 20:53:10.000000000 +0100 -@@ -871,6 +871,10 @@ init_dfa (re_dfa_t *dfa, size_t pat_len) - dfa->state_table = calloc (sizeof (struct re_state_table_entry), table_size); - dfa->state_hash_mask = table_size - 1; - -+#ifndef MB_CUR_MAX -+#define MB_CUR_MAX (0) -+#endif -+ - dfa->mb_cur_max = MB_CUR_MAX; - #ifdef _LIBC - if (dfa->mb_cur_max == 6
package/m4/m4-1.4.9-002-no-wchar.patch deleted +0 -24
@@ -1,24 +0,0 @@ ---- m4-1.4.9/lib/printf-args-orig.h 2007-03-23 05:50:53.000000000 -0700 -+++ m4-1.4.9/lib/printf-args.h 2010-09-18 19:09:21.000000000 -0700 -@@ -29,6 +29,9 @@ - /* Get wint_t. */ - #if HAVE_WINT_T - # include <wchar.h> -+# ifndef __UCLIBC_HAS_WCHAR__ -+typedef int wint_t; -+# endif - #endif - - /* Get va_list. */ ---- m4-1.4.9/lib/wctype_-orig.h 2010-09-18 19:13:07.000000000 -0700 -+++ m4-1.4.9/lib/wctype_.h 2010-09-18 19:09:21.000000000 -0700 -@@ -39,6 +39,9 @@ - # include <stdio.h> - # include <time.h> - # include <wchar.h> -+# ifndef __UCLIBC_HAS_WCHAR__ -+typedef int wint_t; -+# endif - typedef wint_t __wctype_wint_t; - #else - typedef int __wctype_wint_t;
package/m4/m4.mk +1 -5
@@ -3,7 +3,7 @@ # m4 # ############################################################# -M4_VERSION = 1.4.9 +M4_VERSION = 1.4.15 M4_SOURCE = m4-$(M4_VERSION).tar.bz2 M4_SITE = $(BR2_GNU_MIRROR)/m4
@@ -13,10 +13,6 @@ endif M4_CONF_ENV = gl_cv_func_gettimeofday_clobber=no -ifneq ($(BR2_USE_WCHAR),y) -M4_CONF_ENV += gt_cv_c_wchar_t=no gl_cv_absolute_wchar_h=__fpending.h -endif - HOST_M4_CONF_OPT = --disable-static $(eval $(call AUTOTARGETS,package,m4))
Signature · 1 line
1.7.2.3
PATCH 3/3 Added gcc 4.5.1 to buildroot
In reply to Martin Bankympc: gcc 4.5.1 requires this now.
gcc: The patches from gcc 4.4.4 are still needed
--- package/Config.in | 1 + package/mpc/Config.in | 20 +++ package/mpc/mpc.mk | 111 ++++++++++++++++ toolchain/binutils/binutils.mk | 16 +++ toolchain/gcc/4.5.1/100-uclibc-conf.patch | 33 +++++ toolchain/gcc/4.5.1/301-missing-execinfo_h.patch | 11 ++ toolchain/gcc/4.5.1/302-c99-snprintf.patch | 13 ++ .../gcc/4.5.1/305-libmudflap-susv3-legacy.patch | 49 +++++++ toolchain/gcc/4.5.1/810-arm-softfloat-libgcc.patch | 38 ++++++ .../powerpc-link-with-math-lib.patch.conditional | 137 ++++++++++++++++++++ toolchain/gcc/Config.in | 5 + toolchain/gcc/gcc-uclibc-4.x.mk | 14 ++- 12 files changed, 447 insertions(+), 1 deletions(-) create mode 100644 package/mpc/Config.in create mode 100644 package/mpc/mpc.mk create mode 100644 toolchain/gcc/4.5.1/100-uclibc-conf.patch create mode 100644 toolchain/gcc/4.5.1/301-missing-execinfo_h.patch create mode 100644 toolchain/gcc/4.5.1/302-c99-snprintf.patch create mode 100644 toolchain/gcc/4.5.1/305-libmudflap-susv3-legacy.patch create mode 100644 toolchain/gcc/4.5.1/810-arm-softfloat-libgcc.patch create mode 100644 toolchain/gcc/4.5.1/powerpc-link-with-math-lib.patch.conditional
package/Config.in +1 -0
@@ -61,6 +61,7 @@ source "package/grep/Config.in" endif source "package/intltool/Config.in" source "package/make/Config.in" +source "package/mpc/Config.in" source "package/mpfr/Config.in" source "package/libtool/Config.in" source "package/m4/Config.in"
package/mpc/Config.in new file +20 -0
@@ -0,0 +1,20 @@ +config BR2_PACKAGE_LIBMPC + bool "mpc" + select BR2_PACKAGE_LIBMPFR + select BR2_PACKAGE_LIBGMP + help + Mpc is a C library for the arithmetic of complex numbers with + arbitrarily high precision and correct rounding of the result. + It is built upon and follows the same principles as Mpfr. + + http://www.multiprecision.org/ + + +config BR2_PACKAGE_LIBMPC_HEADERS + bool "mpc headers for target" + depends on BR2_PACKAGE_LIBMPC + select BR2_PACKAGE_LIBMPFR_HEADERS + select BR2_PACKAGE_LIBGMP_HEADERS + help + Install the mpc headers for the target. +
package/mpc/mpc.mk new file +111 -0
111 lines not on this page · raw message
toolchain/binutils/binutils.mk +16 -0
@@ -29,6 +29,22 @@ EXTRA_BINUTILS_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_HOST_DIR)" BINUTILS_TARGET_CONFIG_OPTIONS=--with-gmp="$(GMP_TARGET_DIR)" BINUTILS_TARGET_CONFIG_OPTIONS+=--with-mpfr="$(MPFR_TARGET_DIR)" +ifeq ($(BR2_PACKAGE_LIBMPC),y) +BINUTILS_ADD_MPC = y +endif + +ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.) +BINUTILS_ADD_MPC = y +endif + +ifeq ($(BINUTILS_ADD_MPC),y) +BINUTILS_HOST_PREREQ += $(TOOLCHAIN_DIR)/mpc/lib/libmpc$(HOST_LIBEXT) +HOST_SOURCE += host-libmpc-source +BINUTILS_TARGET_PREREQ += $(TARGET_DIR)/usr/lib/libmpc$(LIBTGTEXT) +EXTRA_BINUTILS_CONFIG_OPTIONS += --with-mpc="$(MPC_HOST_DIR)" +BINUTILS_TARGET_CONFIG_OPTIONS += --with-mpc="$(MPC_TARGET_DIR)" +endif + BINUTILS_PATCH_DIR:=toolchain/binutils/$(BINUTILS_VERSION) ifneq ($(filter xtensa%,$(ARCH)),)
toolchain/gcc/4.5.1/100-uclibc-conf.patch new file +33 -0
@@ -0,0 +1,33 @@ +--- gcc/gcc/config/--- gcc/contrib/regression/objs-gcc.sh ++++ gcc/contrib/regression/objs-gcc.sh +@@ -105,6 +105,10 @@ + then + make all-gdb all-dejagnu all-ld || exit 1 + make install-gdb install-dejagnu install-ld || exit 1 ++elif [ $H_REAL_TARGET = $H_REAL_HOST -a $H_REAL_TARGET = i686-pc-linux-uclibc ] ++ then ++ make all-gdb all-dejagnu all-ld || exit 1 ++ make install-gdb install-dejagnu install-ld || exit 1 + elif [ $H_REAL_TARGET = $H_REAL_HOST ] ; then + make bootstrap || exit 1 + make install || exit 1 +--- gcc/libjava/classpath/ltconfig ++++ gcc/libjava/classpath/ltconfig +@@ -603,7 +603,7 @@ + + # Transform linux* to *-*-linux-gnu*, to support old configure scripts. + case $host_os in +-linux-gnu*) ;; ++linux-gnu*|linux-uclibc*) ;; + linux*) host=`echo $host | sed 's/^\(.*-.*-linux\)\(.*\)$/\1-gnu\2/'` + esac + +@@ -1251,7 +1251,7 @@ + ;; + + # This must be Linux ELF. +-linux-gnu*) ++linux*) + version_type=linux + need_lib_prefix=no + need_version=no
toolchain/gcc/4.5.1/301-missing-execinfo_h.patch new file +11 -0
@@ -0,0 +1,11 @@ +--- gcc-4.0.0/boehm-gc/include/gc.h-orig 2005-04-28 22:28:57.000000000 -0500 ++++ gcc-4.0.0/boehm-gc/include/gc.h 2005-04-28 22:30:38.000000000 -0500 +@@ -500,7 +500,7 @@ + #ifdef __linux__ + # include <features.h> + # if (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 1 || __GLIBC__ > 2) \ +- && !defined(__ia64__) ++ && !defined(__ia64__) && !defined(__UCLIBC__) + # ifndef GC_HAVE_BUILTIN_BACKTRACE + # define GC_HAVE_BUILTIN_BACKTRACE + # endif
toolchain/gcc/4.5.1/302-c99-snprintf.patch new file +13 -0
@@ -0,0 +1,13 @@ +Index: gcc-4.3.0/libstdc++-v3/include/c_global/cstdio +=================================================================== +--- gcc-4.3.0/libstdc++-v3/include/c_global/cstdio (revision 129202) ++++ gcc-4.3.0/libstdc++-v3/include/c_global/cstdio (working copy) +@@ -144,7 +144,7 @@ + + _GLIBCXX_END_NAMESPACE + +-#if _GLIBCXX_USE_C99 ++#if _GLIBCXX_USE_C99 || defined __UCLIBC__ + + #undef snprintf + #undef vfscanf
toolchain/gcc/4.5.1/305-libmudflap-susv3-legacy.patch new file +49 -0
49 lines not on this page · raw message
toolchain/gcc/4.5.1/810-arm-softfloat-libgcc.patch new file +38 -0
@@ -0,0 +1,38 @@ +[PATCH] add the correct symbols to libgcc for uclibc arm softfloat + +Signed-off-by: Peter Korsgaard <[email protected]> +--- + gcc/config/arm/linux-elf.h | 2 +- + gcc/config/arm/t-linux | 6 +++++- + 2 files changed, 6 insertions(+), 2 deletions(-) + +Index: gcc-4.4.0/gcc/config/arm/t-linux +=================================================================== +--- gcc-4.4.0.orig/gcc/config/arm/t-linux ++++ gcc-4.4.0/gcc/config/arm/t-linux +@@ -4,7 +4,11 @@ + + LIB1ASMSRC = arm/lib1funcs.asm + LIB1ASMFUNCS = _udivsi3 _divsi3 _umodsi3 _modsi3 _dvmd_lnx _clzsi2 _clzdi2 \ +- _arm_addsubdf3 _arm_addsubsf3 ++ _arm_addsubdf3 _arm_addsubsf3 \ ++ _arm_negdf2 _arm_muldivdf3 _arm_cmpdf2 _arm_unorddf2 \ ++ _arm_fixdfsi _arm_fixunsdfsi _arm_truncdfsf2 \ ++ _arm_negsf2 _arm_muldivsf3 _arm_cmpsf2 _arm_unordsf2 \ ++ _arm_fixsfsi _arm_fixunssfsi + + # MULTILIB_OPTIONS = mhard-float/msoft-float + # MULTILIB_DIRNAMES = hard-float soft-float +Index: gcc-4.4.0/gcc/config/arm/linux-elf.h +=================================================================== +--- gcc-4.4.0.orig/gcc/config/arm/linux-elf.h ++++ gcc-4.4.0/gcc/config/arm/linux-elf.h +@@ -60,7 +60,7 @@ + %{shared:-lc} \ + %{!shared:%{profile:-lc_p}%{!profile:-lc}}" + +-#define LIBGCC_SPEC "%{msoft-float:-lfloat} %{mfloat-abi=soft*:-lfloat} -lgcc" ++#define LIBGCC_SPEC "-lgcc" + + #define GLIBC_DYNAMIC_LINKER "/lib/ld-linux.so.2" +
toolchain/gcc/4.5.1/powerpc-link-with-math-lib.patch.conditional new file +137 -0
137 lines not on this page · raw message
toolchain/gcc/Config.in +5 -0
@@ -25,6 +25,10 @@ choice depends on !BR2_avr32 bool "gcc 4.4.x" + config BR2_GCC_VERSION_4_5_X + depends on !BR2_avr32 + bool "gcc 4.5.x" + config BR2_GCC_VERSION_SNAP depends on !BR2_avr32 bool "gcc snapshot"
@@ -48,6 +52,7 @@ config BR2_GCC_VERSION default "4.2.4" if BR2_GCC_VERSION_4_2_4 default "4.3.5" if BR2_GCC_VERSION_4_3_X default "4.4.4" if BR2_GCC_VERSION_4_4_X + default "4.5.1" if BR2_GCC_VERSION_4_5_X default $BR2_GCC_SNAP_DATE if BR2_GCC_VERSION_SNAP config BR2_EXTRA_GCC_CONFIG_OPTIONS
toolchain/gcc/gcc-uclibc-4.x.mk +13 -1
@@ -101,12 +101,20 @@ GCC_WITH_HOST_GMP=--with-gmp=$(GMP_HOST_DIR) GCC_WITH_HOST_MPFR=--with-mpfr=$(MPFR_HOST_DIR) HOST_SOURCE += host-libgmp-source host-libmpfr-source +ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.) +GCC_WITH_HOST_MPC=--with-mpc=$(MPC_HOST_DIR) +HOST_SOURCE += host-libmpc-source +endif + ifeq ($(BR2_INSTALL_FORTRAN),y) GCC_TARGET_LANGUAGES:=$(GCC_TARGET_LANGUAGES),fortran #GCC_TARGET_PREREQ+=$(TARGET_DIR)/usr/lib/libmpfr.so $(TARGET_DIR)/usr/lib/libgmp.so #GCC_STAGING_PREREQ+=$(TOOLCHAIN_DIR)/mpfr/lib/libmpfr.so GCC_WITH_TARGET_GMP=--with-gmp="$(GMP_TARGET_DIR)" GCC_WITH_TARGET_MPFR=--with-mpfr="$(MPFR_TARGET_DIR)" +ifeq ($(findstring x4.5.,x$(GCC_VERSION)),x4.5.) +GCC_WITH_TARGET_MPC=--with-mpc="$(MPC_TARGET_DIR)" +endif endif ifeq ($(BR2_GCC_SHARED_LIBGCC),y)
@@ -207,6 +215,7 @@ $(GCC_BUILD_DIR1)/.configured: $(GCC_DIR)/.patched $(GCC_TLS) \ $(GCC_WITH_HOST_GMP) \ $(GCC_WITH_HOST_MPFR) \ + $(GCC_WITH_HOST_MPC) \ $(DISABLE_NLS) \ $(THREADS) \ $(GCC_DECIMAL_FLOAT) \
@@ -282,6 +291,7 @@ $(GCC_BUILD_DIR2)/.configured: $(GCC_DIR)/.patched $(GCC_TLS) \ $(GCC_WITH_HOST_GMP) \ $(GCC_WITH_HOST_MPFR) \ + $(GCC_WITH_HOST_MPC) \ $(DISABLE_NLS) \ $(THREADS) \ $(MULTILIB) \
@@ -358,6 +368,7 @@ $(GCC_BUILD_DIR3)/.configured: $(GCC_SRC_DIR)/.patched $(GCC_STAGING_PREREQ) $(GCC_SHARED_LIBGCC) \ $(GCC_WITH_HOST_GMP) \ $(GCC_WITH_HOST_MPFR) \ + $(GCC_WITH_HOST_MPC) \ $(DISABLE_NLS) \ $(THREADS) \ $(GCC_DECIMAL_FLOAT) \
@@ -462,7 +473,7 @@ gcc-dirclean: gcc_initial-dirclean GCC_BUILD_DIR4:=$(BUILD_DIR)/gcc-$(GCC_VERSION)-target $(GCC_BUILD_DIR4)/.prepared: $(STAMP_DIR)/gcc_libs_target_installed $(GCC_TARGET_PREREQ) - mkdir -p $(GCC_BUILD_DIR3) + mkdir -p $(GCC_BUILD_DIR4) touch $@ $(GCC_BUILD_DIR4)/.configured: $(GCC_BUILD_DIR4)/.prepared
@@ -491,6 +502,7 @@ $(GCC_BUILD_DIR4)/.configured: $(GCC_BUILD_DIR4)/.prepared $(GCC_SHARED_LIBGCC) \ $(GCC_WITH_TARGET_GMP) \ $(GCC_WITH_TARGET_MPFR) \ + $(GCC_WITH_TARGET_MPC) \ $(DISABLE_NLS) \ $(THREADS) \ $(GCC_DECIMAL_FLOAT) \
Signature · 1 line
1.7.2.3
Quoted 9 lines · Martin Banky
This is the first pass at adding gcc 4.5.1 to buildroot. It builds fine for the host and the target. The question I have, is the implementation the correct way to change the toolchain? I couldn't find a lot of documentation regarding the toolchain (as far as changing or updating it). Also, can the toolchain mk files be updated to the new format or are their builds too complex?
[PATCH 1/3] dbus: bump to 1.4.0
[PATCH 2/3] m4: bump to 1.4.15
[PATCH 3/3] Added gcc 4.5.1 to buildroot
Hi.
I've been testing this and stuck in a few fixes.
My changes are at git://repo.or.cz/buildroot-gz.git
Fixes are various:
* Fix mpc download site
* Forward-port the powerpc softfloat patch (you probably didn't hit the problem because it's conditional)
* Disable targetopts for powerpc and gcc 4.5.x
* Fix a ton of ARM issues
Word of caution, gcc 4.5 isn't able to build good ARM kernels that can load modules for versions <2.6.36 unless a patch is applied, see
https://bugzilla.kernel.org/show_bug.cgi?id=16310
Tested for arm softfloat (hardware), arm hardfloat (qemu), mips32 (qemu), powerpc hardfloat (e300c3) and powerpc softfloat (qemu). Thanks for your work on this.
Regards.
Gustavo, All,
Quoted 3 lines · On Wednesday 03 November 2010 20:23:08 Gustavo Zacarias wrote:
Word of caution, gcc 4.5 isn't able to build good ARM kernels that can load modules for versions <2.6.36 unless a patch is applied, see
https://bugzilla.kernel.org/show_bug.cgi?id=16310
From what I understand:
- bug was introduced between 2.6.35-rc1 and -rc2: use the gcc option only for powerpc
- the mentioned gcc option was previously applied to all architectures
- the bug has not been fixed on the 2.6.35.y tree
- it was fixed during the 2.6.36 devel cycle
Which leads to the conclusion that:
- 2.6.34 and prior : not impacted
- 2.6.35 and dot-releases : impacted
- 2.6.36 ( and above ;-) ) : not impacted
Sounds reasonable?
Regards,
Yann E. MORIN.
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------. | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | | +33 223 225 172 `------------.-------: X AGAINST | \e/ There is no | | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | '------------------------------^-------^------------------^--------------------'
Quoted 19 lines · Yann E. MORIN
Gustavo, All,
Quoted 3 lines · On Wednesday 03 November 2010 20:23:08 Gustavo Zacarias wrote:
Word of caution, gcc 4.5 isn't able to build good ARM kernels that can load modules for versions <2.6.36 unless a patch is applied, see
https://bugzilla.kernel.org/show_bug.cgi?id=16310From what I understand:
- bug was introduced between 2.6.35-rc1 and -rc2: use the gcc option only for powerpc
- the mentioned gcc option was previously applied to all architectures
- the bug has not been fixed on the 2.6.35.y tree
- it was fixed during the 2.6.36 devel cycleWhich leads to the conclusion that:
- 2.6.34 and prior : not impacted
- 2.6.35 and dot-releases : impacted
- 2.6.36 ( and above ;-) ) : not impacted
Quoted 4 lines
Sounds reasonable?
Regards,
Yann E. MORIN.
See also... http://forums.arm.com/index.php?showtopic=8122
It seems gcc 4.5.x emits R_ARM_REL32 relocations by default unless -fno-dwarf2-cfi-asm is passed as a build option and the ARM kernel doesn't handle it (see arch/arm/include/asm/elf.h and arch/arm/kernel/module.c).
I've Considered patching gcc to disable it by default in buildroot but since it's the kernel at fault i don't see the value in it yet, we'd just be punishing userland in the process too.
I didn't try with <2.6.35 myself, i'll give it a shot tomorrow to see how extended this problem might be (bug report talks about 4.4.x rather than 4.5.x and i've never hit the issue it with 4.4.x). Regards.
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Gustavo" == Gustavo Zacarias <[email protected]> writes:
Gustavo> I've been testing this and stuck in a few fixes.
Gustavo> My changes are at git://repo.or.cz/buildroot-gz.git
Gustavo> Fixes are various:
Gustavo> * Fix mpc download site
Gustavo> * Forward-port the powerpc softfloat patch (you probably didn't hit the Gustavo> problem because it's conditional)
Gustavo> * Disable targetopts for powerpc and gcc 4.5.x
Gustavo> * Fix a ton of ARM issues
Thanks, pulled and pushed.
Signature · 1 line
Bye, Peter Korsgaard