Reproducible builds
Hi,
some time ago, I worked on getting reproducible builds for buildroot. I take "reproducible builds" in a broader sense than what Debian does for instance: in the case of buildroot, we would like the built binaries to be as independent of the build environment as possible, so that in addition to getting identical binaries when building twice on the same machine, we also get identical binaries when building on a different distribution. So far, I got reproducible builds for a (relatively reduced) subset of buildroot packages, on a Debian jessie and Slackware 14.1, with as many differences in the distributions configuration as I could find (timezone, locale, date, time, etc... even the awk flavor used). The only thing I have to get identical is the build path, this is what Debian does, a patch has been proposed a long time ago to allow gcc to avoid the dependence between the __FILE__ macro and the build path, but it does not seem to have been merged:
https://mail-index.netbsd.org/tech-toolchain/2009/02/17/msg000577.html
I expected to be able to clean up the patches before submission (as many patches do the same thing, like overriding the __DATE__ and __TIME__ macros in order to avoid depending on the build time, so this could probably be made generic by buildroot core), but I am not going to have time soon to do that, so here come the patches as they are, in order to spark discussion and gather remarks, before I have time to do more. The patches are based on the 2015.11.1 release.
Regards.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 1/34 reproducibility: introduce config knob
Making the builds reproducible involves turning off build timestamps, which some users may find annoying. So make the reproducible builds optional.
--- Config.in | 3 +++ 1 file changed, 3 insertions(+)
Config.in +3 -0
@@ -623,6 +623,9 @@ config BR2_COMPILER_PARANOID_UNSAFE_PATH endmenu +config BR2_REPRODUCIBLE + bool "Try and remove all sources of non-reproducibility from the build" + endmenu source "toolchain/Config.in"
Signature · 1 line
2.7.4
PATCH 2/34 reproducibility: override locale and timezone
--- Makefile | 7 +++++++ 1 file changed, 7 insertions(+)
Makefile +7 -0
@@ -218,6 +218,13 @@ ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) -include $(BR2_CONFIG) endif +export BR2_REPRODUCIBLE +ifeq ($(BR2_REPRODUCIBLE),y) +export TZ=UTC +export LANG=C +export LC_ALL=C +endif + # To put more focus on warnings, be less verbose as default # Use 'make V=1' to see the full commands ifeq ("$(origin V)", "command line")
Signature · 1 line
2.7.4
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Gilles ChanteperdrixWhen reproducibility is wanted, generates a global SOURCE_DATE_EPOCH environment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.
This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
--- Makefile | 5 +++++ 1 file changed, 5 insertions(+)
Makefile +5 -0
@@ -223,6 +223,11 @@ ifeq ($(BR2_REPRODUCIBLE),y) export TZ=UTC export LANG=C export LC_ALL=C +SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at) +SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s) +export SOURCE_DATE_EPOCH = $(if $(wildcard $(TOPDIR)/.git),$(SOURCE_DATE_GIT),$(SOURCE_DATE_CHANGES)) endif # To put more focus on warnings, be less verbose as default
Signature · 1 line
2.7.4
PATCH 4/34 reproducibility: make rootfs.tar reproducible
In reply to Gilles ChanteperdrixBy generating a tarball with a deterministic file order and date.
--- fs/tar/tar.mk | 11 +++++++++++ 1 file changed, 11 insertions(+)
fs/tar/tar.mk +11 -0
@@ -6,8 +6,19 @@ TAR_OPTS := $(call qstrip,$(BR2_TARGET_ROOTFS_TAR_OPTIONS)) +ifneq ($(BR2_REPRODUCIBLE),y) define ROOTFS_TAR_CMD tar -c$(TAR_OPTS)f $@ --numeric-owner -C $(TARGET_DIR) . endef +else +define ROOTFS_TAR_CMD + cd $(TARGET_DIR) && { \ + find . -\( -! -type d -o -empty -\) -print0 | \ + sort -z | \ + tar --null -T - -c$(TAR_OPTS)f $@ --mtime=@$(SOURCE_DATE_EPOCH) --numeric-owner; \ + } +endef +endif + $(eval $(call ROOTFS_TARGET,tar))
Signature · 1 line
2.7.4
PATCH 5/34 reproducibility/linux: override build timestamp
In reply to Gilles Chanteperdrixwith SOURCE_DATE_EPOCH
--- linux/linux.mk | 8 ++++++++ 1 file changed, 8 insertions(+)
linux/linux.mk +8 -0
@@ -74,6 +74,14 @@ LINUX_MAKE_ENV = \ $(TARGET_MAKE_ENV) \ BR_BINARIES_DIR=$(BINARIES_DIR) +ifeq ($(BR2_REPRODUCIBLE),y) +LINUX_MAKE_ENV += \ + KBUILD_BUILD_VERSION=1 \ + KBUILD_BUILD_USER=buildroot \ + KBUILD_BUILD_HOST=buildroot \ + KBUILD_BUILD_TIMESTAMP="$(shell TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH))" +endif + # Get the real Linux version, which tells us where kernel modules are # going to be installed in the target filesystem. LINUX_VERSION_PROBED = `$(MAKE) $(LINUX_MAKE_FLAGS) -C $(LINUX_DIR) --no-print-directory -s kernelrelease 2>/dev/null`
Signature · 1 line
2.7.4
PATCH 6/34 reproducibility/busybox: disable build timestamps
In reply to Gilles Chanteperdrix--- package/busybox/busybox.mk | 6 ++++++ 1 file changed, 6 insertions(+)
package/busybox/busybox.mk +6 -0
@@ -32,6 +32,12 @@ BUSYBOX_MAKE_ENV = \ $(TARGET_MAKE_ENV) \ CFLAGS="$(BUSYBOX_CFLAGS)" \ CFLAGS_busybox="$(BUSYBOX_CFLAGS_busybox)" + +ifeq ($(BR2_REPRODUCIBLE),y) +BUSYBOX_MAKE_ENV += \ + KCONFIG_NOTIMESTAMP=1 +endif + BUSYBOX_MAKE_OPTS = \ CC="$(TARGET_CC)" \ ARCH=$(KERNEL_ARCH) \
Signature · 1 line
2.7.4
PATCH 7/34 reproducibility/libgcrypt: override timestamps
In reply to Gilles Chanteperdrixwith SOURCE_DATE_EPOCH
--- package/libgcrypt/0003-honor-source-date-epoch.patch | 15 +++++++++++++++ package/libgcrypt/libgcrypt.mk | 1 + 2 files changed, 16 insertions(+) create mode 100644 package/libgcrypt/0003-honor-source-date-epoch.patch
package/libgcrypt/0003-honor-source-date-epoch.patch new file +15 -0
@@ -0,0 +1,15 @@ +--- libgcrypt-1.6.4/configure.ac.orig 2015-09-08 08:27:26.000000000 +0200 ++++ libgcrypt-1.6.4/configure.ac 2016-02-23 21:31:19.290079195 +0100 +@@ -1993,7 +1993,11 @@ changequote([,])dnl + BUILD_FILEVERSION="${BUILD_FILEVERSION}mym4_revision_dec" + AC_SUBST(BUILD_FILEVERSION) + +-BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++if test x"${SOURCE_DATE_EPOCH}" = x; then ++ BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++else ++ BUILD_TIMESTAMP=`date -u -d @${SOURCE_DATE_EPOCH} +%Y-%m-%dT%H:%M+0000 2>/dev/null || date -d @${SOURCE_DATE_EPOCH}` ++fi + AC_SUBST(BUILD_TIMESTAMP) + AC_DEFINE_UNQUOTED(BUILD_TIMESTAMP, "$BUILD_TIMESTAMP", + [The time this package was configured for a build])
package/libgcrypt/libgcrypt.mk +1 -0
@@ -12,6 +12,7 @@ LIBGCRYPT_SITE = ftp://ftp.gnupg.org/gcrypt/libgcrypt LIBGCRYPT_INSTALL_STAGING = YES LIBGCRYPT_DEPENDENCIES = libgpg-error LIBGCRYPT_CONFIG_SCRIPTS = libgcrypt-config +LIBGCRYPT_AUTORECONF = YES LIBGCRYPT_CONF_ENV = \ ac_cv_sys_symbol_underscore=no
Signature · 1 line
2.7.4
PATCH 8/34 reproducibility/libgpg-error: override timestamps
In reply to Gilles Chanteperdrixwith SOURCE_DATE_EPOCH
--- package/libgpg-error/0002-honor-source-date-epoch.patch | 15 +++++++++++++++ 1 file changed, 15 insertions(+) create mode 100644 package/libgpg-error/0002-honor-source-date-epoch.patch
package/libgpg-error/0002-honor-source-date-epoch.patch new file +15 -0
@@ -0,0 +1,15 @@ +--- libgpg-error-1.12/configure.ac~ 2016-02-23 20:30:21.000000000 +0100 ++++ libgpg-error-1.12/configure.ac 2016-02-23 21:16:20.284133994 +0100 +@@ -198,7 +198,11 @@ changequote([,])dnl + BUILD_FILEVERSION="${BUILD_FILEVERSION}0,mym4_revision_dec" + AC_SUBST(BUILD_FILEVERSION) + +-BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++if test x"${SOURCE_DATE_EPOCH}" = x; then ++ BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++else ++ BUILD_TIMESTAMP=`date -u -d @${SOURCE_DATE_EPOCH} +%Y-%m-%dT%H:%M+0000 2>/dev/null || date -d @${SOURCE_DATE_EPOCH}` ++fi + AC_SUBST(BUILD_TIMESTAMP) + AC_DEFINE_UNQUOTED(BUILD_TIMESTAMP, "$BUILD_TIMESTAMP", + [The time this package was configured for a build])
Signature · 1 line
2.7.4
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Gilles Chanteperdrixglib2 installs compiled python bytecode in /usr/share/glib-2.0/codegen, likely for the gdbus-codegen program.
This bytecode is not compiled with the buildroot host python, which causes a first source of non reproducibility, fix that by using buildroot host python.
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries, so, since buildroot users are unlikely to need the qdbus-codegen program on target, add an option to remove it. The option is disabled by default.
--- package/libglib2/Config.in | 10 ++++++++++ package/libglib2/libglib2.mk | 18 ++++++++++++++++-- 2 files changed, 26 insertions(+), 2 deletions(-)
package/libglib2/Config.in +10 -0
@@ -13,6 +13,16 @@ config BR2_PACKAGE_LIBGLIB2 http://www.gtk.org/ +if BR2_PACKAGE_LIBGLIB2 + +config BR2_PACKAGE_LIBGLIB2_CODEGEN + bool "install Glib D-Bus code and documentation generator" + default y + help + Install Glib D-Bus code and documentation generator. + +endif + comment "libglib2 needs a toolchain w/ wchar, threads" depends on BR2_USE_MMU depends on !BR2_USE_WCHAR || !BR2_TOOLCHAIN_HAS_THREADS
package/libglib2/libglib2.mk +16 -2
@@ -99,9 +99,9 @@ HOST_LIBGLIB2_CONF_OPTS = \ --disable-systemtap \ --disable-xattr -LIBGLIB2_DEPENDENCIES = host-pkgconf host-libglib2 libffi zlib $(if $(BR2_NEEDS_GETTEXT),gettext) host-gettext +LIBGLIB2_DEPENDENCIES = host-pkgconf host-libglib2 libffi zlib $(if $(BR2_NEEDS_GETTEXT),gettext) host-gettext host-python -HOST_LIBGLIB2_DEPENDENCIES = host-pkgconf host-libffi host-zlib host-gettext +HOST_LIBGLIB2_DEPENDENCIES = host-pkgconf host-libffi host-zlib host-gettext host-python ifneq ($(BR2_ENABLE_LOCALE),y) LIBGLIB2_DEPENDENCIES += libiconv
@@ -137,6 +137,20 @@ ifneq ($(BR2_PACKAGE_GDB),y) LIBGLIB2_POST_INSTALL_TARGET_HOOKS += LIBGLIB2_REMOVE_GDB_FILES endif +ifeq ($(BR2_PACKAGE_LIBGLIB2_CODEGEN),) +define LIBGLIB2_REMOVE_CODEGEN + rm $(TARGET_DIR)/usr/bin/gdbus-codegen + rm -rf $(TARGET_DIR)/usr/share/glib-2.0/codegen + rmdir --ignore-fail-on-non-empty $(TARGET_DIR)/usr/share/glib-2.0 +endef + +LIBGLIB2_POST_INSTALL_TARGET_HOOKS += LIBGLIB2_REMOVE_CODEGEN +endif + +LIBGLIB2_CONF_OPTS += --with-python=$(HOST_DIR)/usr/bin/python + +HOST_LIBGLIB2_CONF_OPTS += --with-python=$(HOST_DIR)/usr/bin/python + $(eval $(autotools-package)) $(eval $(host-autotools-package))
Signature · 1 line
2.7.4
PATCH 10/34 reproducibility/sysstat: override gcc macros
In reply to Gilles ChanteperdrixOverride the __DATE__ and __TIME__ macros with values obtained from the package SOURCE_DATE_EPOCH.
--- package/sysstat/sysstat.mk | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
package/sysstat/sysstat.mk +19 -0
@@ -17,6 +17,25 @@ SYSSTAT_DEPENDENCIES += gettext SYSSTAT_MAKE_OPTS += LFLAGS="$(TARGET_LDFLAGS) -lintl" endif +SYSSTAT_CONF_OPTS += \ + sa_lib_dir=/usr/lib/sa + +ifeq ($(BR2_REPRODUCIBLE),y) +SYSSTAT_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +SYSSTAT_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +SYSSTAT_CONF_OPTS += \ + CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(SYSSTAT_DIR)/.br_date_override.h" + +define SYSSTAT_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(SYSSTAT_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(SYSSTAT_SOURCE_DATE))" >> $(SYSSTAT_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(SYSSTAT_SOURCE_TIME))" >> $(SYSSTAT_DIR)/.br_date_override.h +endef + +SYSSTAT_POST_PATCH_HOOKS += SYSSTAT_OVERRIDE_DATE_HOOK +endif + # The isag tool is a post processing script that depends on tcl/tk # among other things. So we don't install it. SYSSTAT_INSTALL_TARGET_OPTS = DESTDIR=$(TARGET_DIR) INSTALL_ISAG=n install
Signature · 1 line
2.7.4
PATCH 11/34 reproducibility/vim: make vim reproducible
In reply to Gilles ChanteperdrixThere are three reasons why vim is not reproducible:
- it embeds the build timestamp, fix that by overriding them with SOURCE_DATE_EPOCH;
- it embeds the build host and user, fix that by using its configure script --with-compiledby option
- it generates an awk script which depends on the version of awk present on the build machine, fix that by forcing the interpreter to be /usr/bin/awk, it is what is available on buildroot rootfs anyway.
--- package/vim/vim.mk | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+)
package/vim/vim.mk +23 -0
@@ -34,6 +34,29 @@ else VIM_CONF_OPTS += --disable-selinux endif +ifeq ($(BR2_REPRODUCIBLE),y) +VIM_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +VIM_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +VIM_CONF_OPTS += \ + --with-compiledby=buildroot \ + CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(VIM_DIR)/.br_date_override.h" + +define VIM_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(VIM_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(VIM_SOURCE_DATE))" >> $(VIM_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(VIM_SOURCE_TIME))" >> $(VIM_DIR)/.br_date_override.h +endef + +VIM_POST_PATCH_HOOKS += VIM_OVERRIDE_DATE_HOOK + +define VIM_FIX_AWK_REFERENCE + sed -i 's,^#!.*awk,#!/usr/bin/awk,' $(TARGET_DIR)/usr/share/vim/vim74/tools/mve.awk +endef + +VIM_POST_INSTALL_TARGET_HOOKS += VIM_FIX_AWK_REFERENCE +endif + define VIM_INSTALL_TARGET_CMDS cd $(@D)/src; \ $(MAKE) DESTDIR=$(TARGET_DIR) installvimbin; \
Signature · 1 line
2.7.4
PATCH 12/34 reproducibility/wireshark: remove documentation
In reply to Gilles ChanteperdrixHTML documentation contains a link to a mail address specific to the distribution of the build machine. Remove html documentation from the rootfs as it is unlikely to be needed. At this chance also remove the txt documentation.
--- package/wireshark/wireshark.mk | 6 ++++++ 1 file changed, 6 insertions(+)
package/wireshark/wireshark.mk +6 -0
@@ -28,4 +28,10 @@ WIRESHARK_CONF_OPTS = \ --with-libsmi=no \ --includedir=$(STAGING_DIR)/usr/include +define WIRESHARK_REMOVE_DOCS + find $(TARGET_DIR)/usr/share/wireshark -\( -name '*.txt' -o -name '*.html' -\) -delete +endef + +WIRESHARK_POST_INSTALL_TARGET_HOOKS += WIRESHARK_REMOVE_DOCS + $(eval $(autotools-package))
Signature · 1 line
2.7.4
PATCH 13/34 reproducibility/libtool: fix reproducibility issues
In reply to Gilles Chanteperdrix- remove libtool and libtoolize; they are full of references to the build environment tools, and unlikely to be needed on target;
- do not use /etc/ld.so.conf contents in libtool search path: they are not relevant on target, and cause a useless dependency of libltdl on the build environment.
--- package/libtool/0001-fix-cross-compilation.patch | 11 +++++++++++ package/libtool/libtool.mk | 8 ++++++-- 2 files changed, 17 insertions(+), 2 deletions(-) create mode 100644 package/libtool/0001-fix-cross-compilation.patch
package/libtool/0001-fix-cross-compilation.patch new file +11 -0
@@ -0,0 +1,11 @@ +--- libtool-2.4.6/m4/libtool.m4~ 2015-01-20 17:15:19.000000000 +0100 ++++ libtool-2.4.6/m4/libtool.m4 2016-02-27 21:02:21.513684396 +0100 +@@ -2873,7 +2873,7 @@ linux* | k*bsd*-gnu | kopensolaris*-gnu + # 'ldconfig -N -X -v | grep ^/' on 64bit Fedora does not report /usr/lib64, + # even though it is searched at run-time. Try to do the best guess by + # appending ld.so.conf contents (and includes) to the search path. +- if test -f /etc/ld.so.conf; then ++ if test "$cross_compiling" != "yes" -a -f /etc/ld.so.conf; then + lt_ld_extra=`awk '/^include / { system(sprintf("cd /etc; cat %s 2>/dev/null", \[$]2)); skip = 1; } { if (!skip) print \[$]0; skip = 0; }' < /etc/ld.so.conf | $SED -e 's/#.*//;/^[ ]*hwcap[ ]/d;s/[:, ]/ /g;s/=[^=]*$//;s/=[^= ]* / /g;s/"//g;/^$/d' | tr '\n' ' '` + sys_lib_dlsearch_path_spec="/lib /usr/lib $lt_ld_extra" + fi
package/libtool/libtool.mk +6 -2
@@ -8,7 +8,7 @@ LIBTOOL_VERSION = 2.4.6 LIBTOOL_SOURCE = libtool-$(LIBTOOL_VERSION).tar.xz LIBTOOL_SITE = $(BR2_GNU_MIRROR)/libtool LIBTOOL_INSTALL_STAGING = YES -LIBTOOL_CONF_ENV = HELP2MAN=true +LIBTOOL_CONF_OPTS = HELP2MAN=true LIBTOOL_DEPENDENCIES = host-m4 LIBTOOL_LICENSE = GPLv2+ LIBTOOL_LICENSE_FILES = COPYING
@@ -33,9 +33,13 @@ define LIBTOOL_AVOID_AUTORECONF_HOOK find $(@D) -name configure -exec touch '{}' \; find $(@D) -name Makefile.in -exec touch '{}' \; endef -LIBTOOL_PRE_CONFIGURE_HOOKS += LIBTOOL_AVOID_AUTORECONF_HOOK HOST_LIBTOOL_PRE_CONFIGURE_HOOKS += LIBTOOL_AVOID_AUTORECONF_HOOK +define LIBTOOL_REMOVE_LIBTOOL + rm $(TARGET_DIR)/usr/bin/libtool $(TARGET_DIR)/usr/bin/libtoolize +endef +LIBTOOL_POST_INSTALL_TARGET_HOOKS += LIBTOOL_REMOVE_LIBTOOL + $(eval $(autotools-package)) $(eval $(host-autotools-package))
Signature · 1 line
2.7.4
PATCH 14/34 reproducibility/python: also remove pyo files
In reply to Gilles Chanteperdrixwhen BR2_PACKAGE_PYTHON_PY_ONLY is selected. Make that option the default when BR2_REPRODUCIBLE is selected in order to avoid problems with timestamps embedded in compiled python files.
--- package/python/Config.in | 3 ++- package/python/python.mk | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-)
package/python/Config.in +2 -1
@@ -19,7 +19,8 @@ if BR2_PACKAGE_PYTHON choice prompt "python module format to install" - default BR2_PACKAGE_PYTHON_PYC_ONLY + default BR2_PACKAGE_PYTHON_PYC_ONLY if !BR2_REPRODUCIBLE + default BR2_PACKAGE_PYTHON_PY_ONLY if BR2_REPRODUCIBLE help Select Python module format to install on target (py, pyc or both)
package/python/python.mk +1 -1
@@ -225,7 +225,7 @@ endif ifeq ($(BR2_PACKAGE_PYTHON_PY_ONLY),y) define PYTHON_FINALIZE_TARGET - find $(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR) -name '*.pyc' -print0 | xargs -0 rm -f + find $(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR) -name '*.py[co]' -print0 | xargs -0 rm -f endef endif
Signature · 1 line
2.7.4
PATCH 15/34 reproducibility/gzip: avoid gzip timestamp in rootfs.tar.gz
In reply to Gilles Chanteperdrix--- fs/common.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/common.mk +1 -1
@@ -48,7 +48,7 @@ ROOTFS_$(2)_DEPENDENCIES += host-fakeroot host-makedevs \ ifeq ($$(BR2_TARGET_ROOTFS_$(2)_GZIP),y) ROOTFS_$(2)_COMPRESS_EXT = .gz -ROOTFS_$(2)_COMPRESS_CMD = gzip -9 -c +ROOTFS_$(2)_COMPRESS_CMD = gzip -9 -c $(if $(BR2_REPRODUCIBLE),-n) endif ifeq ($$(BR2_TARGET_ROOTFS_$(2)_BZIP2),y) ROOTFS_$(2)_COMPRESS_EXT = .bz2
Signature · 1 line
2.7.4
PATCH 16/34 reproducibility/lzo: forbid lzo with reproducible builds
In reply to Gilles ChanteperdrixThe lzo archive contains a timestamp.
--- fs/tar/Config.in | 7 +++++++ 1 file changed, 7 insertions(+)
fs/tar/Config.in +7 -0
@@ -33,6 +33,7 @@ config BR2_TARGET_ROOTFS_TAR_LZMA config BR2_TARGET_ROOTFS_TAR_LZO bool "lzo" + depends on !BR2_REPRODUCIBLE help Do compress the tarball with lzop.
@@ -43,6 +44,12 @@ config BR2_TARGET_ROOTFS_TAR_XZ endchoice +if BR2_REPRODUCIBLE + +comment "lzo compression is only available with non-reproducible builds" + +endif + config BR2_TARGET_ROOTFS_TAR_OPTIONS string "other random options to pass to tar" depends on BR2_TARGET_ROOTFS_TAR
Signature · 1 line
2.7.4
PATCH 17/34 reproducibility/libnspr: override build timestamps
In reply to Gilles Chanteperdrixwith SOURCE_DATE_EPOCH
--- package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch | 71 ++++++++++++++++++++++ 1 file changed, 71 insertions(+) create mode 100644 package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch
package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch new file +71 -0
@@ -0,0 +1,71 @@ +--- libnspr-4.10.10/nspr/lib/libc/src/Makefile.in~ 2015-10-16 15:22:19.000000000 +0200 ++++ libnspr-4.10.10/nspr/lib/libc/src/Makefile.in 2016-02-27 16:11:13.745483215 +0100 +@@ -103,7 +103,7 @@ ECHO = echo + TINC = $(OBJDIR)/_pl_bld.h + PROD = $(notdir $(SHARED_LIBRARY)) + NOW = $(MOD_DEPTH)/config/$(OBJDIR)/now +-SH_DATE = $(shell date "+%Y-%m-%d %T") ++SH_DATE = $(shell date `test -n "$$SOURCE_DATE_EPOCH" && echo -d @$$SOURCE_DATE_EPOCH` "+%Y-%m-%d %T") + SH_NOW = $(shell $(NOW)) + + ifeq ($(NS_USE_GCC)_$(OS_ARCH),_WINNT) +--- libnspr-4.10.10/nspr/lib/ds/Makefile.in~ 2015-10-16 15:22:19.000000000 +0200 ++++ libnspr-4.10.10/nspr/lib/ds/Makefile.in 2016-02-27 16:11:28.115738828 +0100 +@@ -101,7 +101,7 @@ ECHO = echo + TINC = $(OBJDIR)/_pl_bld.h + PROD = $(notdir $(SHARED_LIBRARY)) + NOW = $(MOD_DEPTH)/config/$(OBJDIR)/now +-SH_DATE = $(shell date "+%Y-%m-%d %T") ++SH_DATE = $(shell date `test -n "$$SOURCE_DATE_EPOCH" && echo -d @$$SOURCE_DATE_EPOCH` "+%Y-%m-%d %T") + SH_NOW = $(shell $(NOW)) + + ifeq ($(NS_USE_GCC)_$(OS_ARCH),_WINNT) +--- libnspr-4.10.10/nspr/lib/prstreams/Makefile.in~ 2015-10-16 15:22:19.000000000 +0200 ++++ libnspr-4.10.10/nspr/lib/prstreams/Makefile.in 2016-02-27 16:11:39.799279987 +0100 +@@ -105,7 +105,7 @@ ECHO = echo + TINC = $(OBJDIR)/_pl_bld.h + PROD = $(notdir $(SHARED_LIBRARY)) + NOW = $(MOD_DEPTH)/config/$(OBJDIR)/now +-SH_DATE = $(shell date "+%Y-%m-%d %T") ++SH_DATE = $(shell date `test -n "$$SOURCE_DATE_EPOCH" && echo -d @$$SOURCE_DATE_EPOCH` "+%Y-%m-%d %T") + SH_NOW = $(shell $(NOW)) + + ifeq ($(OS_ARCH), WINNT) +--- libnspr-4.10.10/nspr/pr/src/Makefile.in~ 2015-10-16 15:22:19.000000000 +0200 ++++ libnspr-4.10.10/nspr/pr/src/Makefile.in 2016-02-27 16:18:55.560364508 +0100 +@@ -315,7 +315,7 @@ PROD = $(notdir $(SHARED_LIBRARY)) + endif + + NOW = $(MOD_DEPTH)/config/$(OBJDIR)/now +-SH_DATE = $(shell date "+%Y-%m-%d %T") ++SH_DATE = $(shell date `test -n "$$SOURCE_DATE_EPOCH" && echo -d @$$SOURCE_DATE_EPOCH` "+%Y-%m-%d %T") + SH_NOW = $(shell $(NOW)) + + ifeq ($(NS_USE_GCC)_$(OS_ARCH),_WINNT) +--- libnspr-4.10.10/nspr/config/now.c~ 2015-10-16 15:22:19.000000000 +0200 ++++ libnspr-4.10.10/nspr/config/now.c 2016-02-27 16:29:53.862069883 +0100 +@@ -4,6 +4,7 @@ + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ + + #include <stdio.h> ++#include <stdlib.h> + #include <time.h> + + int main(int argc, char **argv) +@@ -27,10 +28,15 @@ int main(int argc, char **argv) + now = (1000000i64) * sec; + fprintf(stdout, "%I64d", now); + #else ++ const char *source_date_epoch = getenv("SOURCE_DATE_EPOCH"); + long long now; + time_t sec; + +- sec = time(NULL); ++ if (!source_date_epoch) ++ sec = time(NULL); ++ else ++ sec = (time_t)atoll(source_date_epoch); ++ + now = (1000000LL) * sec; + fprintf(stdout, "%lld", now); + #endif
Signature · 1 line
2.7.4
PATCH 18/34 readline: fix installed libraries permission
In reply to Gilles ChanteperdrixIn order to avoid strip failing to strip the library.
--- package/readline/0009-avoid-strip-perm-issue.patch | 11 +++++++++++ 1 file changed, 11 insertions(+) create mode 100644 package/readline/0009-avoid-strip-perm-issue.patch
package/readline/0009-avoid-strip-perm-issue.patch new file +11 -0
@@ -0,0 +1,11 @@ +--- readline-6.3/support/shlib-install~ 2012-07-10 21:58:08.000000000 +0200 ++++ readline-6.3/support/shlib-install 2016-02-27 01:19:09.878310797 +0100 +@@ -73,7 +73,7 @@ fi + case "$host_os" in + hpux*|darwin*|macosx*|linux*|solaris2*) + if [ -z "$uninstall" ]; then +- chmod 555 ${INSTALLDIR}/${LIBNAME} ++ chmod 755 ${INSTALLDIR}/${LIBNAME} + fi ;; + cygwin*|mingw*) + IMPLIBNAME=`echo ${LIBNAME} \
Signature · 1 line
2.7.4
PATCH 19/34 reproducibility/python: make python reproducible
In reply to Gilles Chanteperdrixpython build is not reproducible for two reasons:
- the libpython library embeds the build timestamps, fix this by overriding the __DATE__ and __TIME__ macro using SOURCE_DATE_EPOCH;
- _sysconfigdata.py and config/Makefile on target includes references through INSTALL and MKDIR_P variables to host tools, fix this by overriding these variables values with the corresponding paths on buildroot rootfs.
--- package/python/python.mk | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-)
package/python/python.mk +19 -1
@@ -143,7 +143,9 @@ PYTHON_CONF_OPTS += \ --disable-tk \ --disable-nis \ --disable-dbm \ - --disable-pyo-build + --disable-pyo-build \ + ac_cv_path_install=/usr/bin/install \ + ac_cv_path_mkdir=/bin/mkdir # This is needed to make sure the Python build process doesn't try to # regenerate those files with the pgen program. Otherwise, it builds
@@ -214,6 +216,22 @@ endif # Provided to other packages PYTHON_PATH = $(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR)/sysconfigdata/:$(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR)/site-packages/ +ifeq ($(BR2_REPRODUCIBLE),y) +PYTHON_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +PYTHON_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +PYTHON_CONF_OPTS += \ + CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(PYTHON_DIR)/.br_date_override.h" + +define PYTHON_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(PYTHON_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(PYTHON_SOURCE_DATE))" >> $(PYTHON_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(PYTHON_SOURCE_TIME))" >> $(PYTHON_DIR)/.br_date_override.h +endef + +PYTHON_POST_PATCH_HOOKS += PYTHON_OVERRIDE_DATE_HOOK +endif + $(eval $(autotools-package)) $(eval $(host-autotools-package))
Signature · 1 line
2.7.4
PATCH 20/34 reproducibility/iproute2: do not rely on host directories
In reply to Gilles ChanteperdrixIn order to find whether iptables/xtables are installed on target
--- .../0003-avoid-checking-host-directories.patch | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) create mode 100644 package/iproute2/0003-avoid-checking-host-directories.patch
package/iproute2/0003-avoid-checking-host-directories.patch new file +19 -0
@@ -0,0 +1,19 @@ +--- iproute2-4.3.0/configure~ 2016-02-27 20:31:57.000000000 +0100 ++++ iproute2-4.3.0/configure 2016-02-27 20:32:05.301478415 +0100 +@@ -184,16 +184,6 @@ check_ipt_lib_dir() + return + fi + +- for dir in /lib /usr/lib /usr/local/lib +- do +- for file in $dir/{xtables,iptables}/lib*t_*so ; do +- if [ -f $file ]; then +- echo ${file%/*} +- echo "IPT_LIB_DIR:=${file%/*}" >> Config +- return +- fi +- done +- done + echo "not found!" + } +
Signature · 1 line
2.7.4
PATCH 21/34 hddtemp: add
In reply to Gilles Chanteperdrix--- package/Config.in | 1 + package/hddtemp/Config.in | 8 ++++++++ package/hddtemp/hddtemp.hash | 2 ++ package/hddtemp/hddtemp.mk | 19 +++++++++++++++++++ 4 files changed, 30 insertions(+) create mode 100644 package/hddtemp/Config.in create mode 100644 package/hddtemp/hddtemp.hash create mode 100644 package/hddtemp/hddtemp.mk
package/Config.in +1 -0
@@ -1458,6 +1458,7 @@ comment "Utilities" source "package/file/Config.in" source "package/gnupg/Config.in" source "package/gnupg2/Config.in" + source "package/hddtemp/Config.in" source "package/inotify-tools/Config.in" source "package/lockfile-progs/Config.in" source "package/logrotate/Config.in"
package/hddtemp/Config.in new file +8 -0
@@ -0,0 +1,8 @@ +config BR2_PACKAGE_HDDTEMP + bool "hddtemp" + help + hddtemp is a small utility that gives you the temperature of your + hard drive by reading S.M.A.R.T. informations, for drives that + support this feature. + + http://www.guzu.net/linux/hddtemp.php
package/hddtemp/hddtemp.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 618541584054093d53be8a2d9e81c97174f30f00af91cb8700a97e442d79ef5b hddtemp-0.3-beta15.tar.bz2
package/hddtemp/hddtemp.mk new file +19 -0
@@ -0,0 +1,19 @@ +################################################################################ +# +# hddtemp +# +################################################################################ + +HDDTEMP_VERSION = 0.3-beta15 +HDDTEMP_SOURCE = hddtemp-$(HDDTEMP_VERSION).tar.bz2 +HDDTEMP_SITE = http://download.savannah.gnu.org/releases/hddtemp +HDDTEMP_LICENSE = GPLv2 +HDDTEMP_LICENSE_FILES = GPL-2 + +define HDDTEMP_REMOVE_MAN + rm -rf $(TARGET_DIR)/usr/man +endef + +HDDTEMP_POST_INSTALL_TARGET_HOOKS += HDDTEMP_REMOVE_MAN + +$(eval $(autotools-package))
Signature · 1 line
2.7.4
PATCH 22/34 High availability: add packages
In reply to Gilles Chanteperdrixcluster-glue
corosync
heartbeat
libqb
pacemaker
--- package/Config.in | 8 ++ package/cluster-glue/0001-fix-cross-compile.patch | 26 +++++ package/cluster-glue/Config.in | 9 ++ package/cluster-glue/cluster-glue.hash | 2 + package/cluster-glue/cluster-glue.mk | 23 +++++ package/corosync/Config.in | 8 ++ package/corosync/corosync.hash | 2 + package/corosync/corosync.mk | 17 ++++ package/heartbeat/0001-fix-cross-compilation.patch | 107 +++++++++++++++++++++ .../heartbeat/0002-avoid-strip-perm-issue.patch | 10 ++ package/heartbeat/Config.in | 10 ++ package/heartbeat/heartbeat.hash | 2 + package/heartbeat/heartbeat.mk | 37 +++++++ package/libqb/Config.in | 11 +++ package/libqb/libqb.hash | 2 + package/libqb/libqb.mk | 20 ++++ package/pacemaker/Config.in | 9 ++ package/pacemaker/pacemaker.hash | 2 + package/pacemaker/pacemaker.mk | 32 ++++++ 19 files changed, 337 insertions(+) create mode 100644 package/cluster-glue/0001-fix-cross-compile.patch create mode 100644 package/cluster-glue/Config.in create mode 100644 package/cluster-glue/cluster-glue.hash create mode 100644 package/cluster-glue/cluster-glue.mk create mode 100644 package/corosync/Config.in create mode 100644 package/corosync/corosync.hash create mode 100644 package/corosync/corosync.mk create mode 100644 package/heartbeat/0001-fix-cross-compilation.patch create mode 100644 package/heartbeat/0002-avoid-strip-perm-issue.patch create mode 100644 package/heartbeat/Config.in create mode 100644 package/heartbeat/heartbeat.hash create mode 100644 package/heartbeat/heartbeat.mk create mode 100644 package/libqb/Config.in create mode 100644 package/libqb/libqb.hash create mode 100644 package/libqb/libqb.mk create mode 100644 package/pacemaker/Config.in create mode 100644 package/pacemaker/pacemaker.hash create mode 100644 package/pacemaker/pacemaker.mk
package/Config.in +8 -0
@@ -902,6 +902,14 @@ menu "Hardware handling" source "package/urg/Config.in" endmenu +menu "High availability" + source "package/cluster-glue/Config.in" + source "package/corosync/Config.in" + source "package/heartbeat/Config.in" + source "package/libqb/Config.in" + source "package/pacemaker/Config.in" +endmenu + menu "Javascript" source "package/angularjs/Config.in" source "package/bootstrap/Config.in"
package/cluster-glue/0001-fix-cross-compile.patch new file +26 -0
@@ -0,0 +1,26 @@ +--- cluster-glue-0a7add1d9996/configure.ac~ 2014-10-16 19:24:17.000000000 +0200 ++++ cluster-glue-0a7add1d9996/configure.ac 2016-02-26 15:39:02.732961215 +0100 +@@ -231,23 +231,6 @@ esac + AC_SUBST(INITDIR) + + AC_MSG_NOTICE(Sanitizing libdir: ${libdir}) +-case $libdir in +- dnl For consistency with Heartbeat, map NONE->$prefix +- *prefix*|NONE) +- AC_MSG_CHECKING(which lib directory to use) +- for aDir in lib64 lib +- do +- trydir="${exec_prefix}/${aDir}" +- if +- test -d ${trydir} +- then +- libdir=${trydir} +- break +- fi +- done +- AC_MSG_RESULT($libdir); +- ;; +-esac + + DLOPEN_FORCE_FLAGS="" + AC_SUBST(DLOPEN_FORCE_FLAGS)
package/cluster-glue/Config.in new file +9 -0
@@ -0,0 +1,9 @@ +config BR2_PACKAGE_CLUSTER_GLUE + bool "cluster-glue" + select BR2_PACKAGE_UTIL_LINUX + select BR2_PACKAGE_UTIL_LINUX_LIBUUID + help + Cluster Glue is a set of libraries, tools and utilities suitable for + the Heartbeat/Pacemaker cluster stack. + + http://www.linux-ha.org/wiki/Cluster_Glue
package/cluster-glue/cluster-glue.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 c47bfffbfe11401379d5248fc7518a3de9532c02090d614f314037fc3a7c4ceb 0a7add1d9996.tar.bz2
package/cluster-glue/cluster-glue.mk new file +23 -0
@@ -0,0 +1,23 @@ +################################################################################ +# +# cluster-glue +# +################################################################################ + +CLUSTER_GLUE_VERSION = 0a7add1d9996 +CLUSTER_GLUE_SOURCE = $(CLUSTER_GLUE_VERSION).tar.bz2 +CLUSTER_GLUE_SITE = http://hg.linux-ha.org/glue/archive +CLUSTER_GLUE_LICENSE = LGPLv2.1+ (library), GPLv2+ (programs) +CLUSTER_GLUE_LICENSE_FILES = COPYING COPYING.LIB +CLUSTER_GLUE_DEPENDENCIES += host-pkgconf host-python libtool util-linux bzip2 libglib2 libxml2 + +CLUSTER_GLUE_INSTALL_STAGING = YES +CLUSTER_GLUE_AUTORECONF = YES + +CLUSTER_GLUE_CONF_ENV += PYTHON=$(HOST_DIR)/usr/bin/python + +CLUSTER_GLUE_CONF_OPTS += \ + --with-initdir=/etc/init.d \ + ac_cv_path_XML2CONFIG="$(STAGING_DIR)/usr/bin/xml2-config" + +$(eval $(autotools-package))
package/corosync/Config.in new file +8 -0
@@ -0,0 +1,8 @@ +config BR2_PACKAGE_COROSYNC + bool "corosync" + help + The Corosync Cluster Engine is a Group Communication System with + additional features for implementing high availability within + applications. + + https://corosync.github.io/corosync/
package/corosync/corosync.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 64ee50f783a4bfa8a63fd576090cf7dd1810f3c60e01b4886723b87c5ff2b013 corosync-1.4.8.tar.gz
package/corosync/corosync.mk new file +17 -0
@@ -0,0 +1,17 @@ +################################################################################ +# +# corosync +# +################################################################################ + +COROSYNC_VERSION = 1.4.8 +COROSYNC_SOURCE = corosync-$(COROSYNC_VERSION).tar.gz +COROSYNC_SITE = http://build.clusterlabs.org/corosync/releases +COROSYNC_LICENSE = BSD-3c except for crypto.[ch], public domain for crypto.[ch] +COROSYNC_LICENSE_FILES = LICENSE +COROSYNC_INSTALL_STAGING = YES +COROSYNC_DEPENDENCIES += host-pkgconf libnss + +COROSYNC_CONF_OPTS += ac_cv_func_fnmatch_works=yes + +$(eval $(autotools-package))
package/heartbeat/0001-fix-cross-compilation.patch new file +107 -0
107 lines not on this page · raw message
package/heartbeat/0002-avoid-strip-perm-issue.patch new file +10 -0
@@ -0,0 +1,10 @@ +--- Heartbeat-3-0-958e11be8686/tools/Makefile.am~ 2015-02-09 17:04:41.000000000 +0100 ++++ Heartbeat-3-0-958e11be8686/tools/Makefile.am 2016-02-27 01:13:40.102433992 +0100 +@@ -49,6 +49,6 @@ cl_respawn_LDADD = -lplumb \ + + install-data-hook: # install-exec-hook doesn't work (!) + -chgrp $(apigid) $(DESTDIR)/$(habindir)/cl_status +- -chmod g+s,a-w $(DESTDIR)/$(habindir)/cl_status ++ -chmod g+s,g-w,o-w $(DESTDIR)/$(habindir)/cl_status + + .PHONY: install-exec-hook
package/heartbeat/Config.in new file +10 -0
@@ -0,0 +1,10 @@ +config BR2_PACKAGE_HEARTBEAT + bool "heartbeat" + select BR2_PACKAGE_UTIL_LINUX + select BR2_PACKAGE_UTIL_LINUX_LIBUUID + select BR2_PACKAGE_IPROUTE2 + help + Heartbeat is a daemon that provides cluster infrastructure + (communication and membership) services to its clients. + + http://www.linux-ha.org/wiki/Heartbeat
package/heartbeat/heartbeat.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 851d2add2c129fef9fede764fec80229e1f6e7295e0e979950d10258648b462c 958e11be8686.tar.bz2
package/heartbeat/heartbeat.mk new file +37 -0
@@ -0,0 +1,37 @@ +################################################################################ +# +# heartbeat +# +################################################################################ +HEARTBEAT_VERSION = 958e11be8686 +HEARTBEAT_SOURCE = $(HEARTBEAT_VERSION).tar.bz2 +HEARTBEAT_SITE = http://hg.linux-ha.org/heartbeat-STABLE_3_0/archive +HEARTBEAT_LICENSE = LGPLv2.1+ (library), GPLv2+ (programs) +HEARTBEAT_LICENSE_FILES = doc/COPYING doc/COPYING.LGPL +HEARTBEAT_DEPENDENCIES += host-pkgconf host-python libtool util-linux bzip2 cluster-glue +HEARTBEAT_INSTALL_STAGING = YES +HEARTBEAT_AUTORECONF = YES + +HEARTBEAT_CONF_ENV += PYTHON=$(HOST_DIR)/usr/bin/python + +HEARTBEAT_CONF_OPTS += \ + CPPFLAGS="$(TARGET_CPPFLAGS) -I$(STAGING_DIR)/usr/include/heartbeat -I$(STAGING_DIR)/usr/include/clplumbing" \ + --with-initdir=/etc/init.d \ + --with-initdefdir=/etc/sysconfig \ + ac_cv_path_IP2UTIL=/sbin/ip + +ifeq ($(BR2_PACKAGE_PYTHON_PYC_ONLY),y) +define HEARTBEAT_FINALIZE_TARGET + find $(TARGET_DIR)/usr/share/heartbeat -name '*.py' -print0 | xargs -0 rm -f +endef +endif + +ifeq ($(BR2_PACKAGE_PYTHON_PY_ONLY),y) +define HEARTBEAT_FINALIZE_TARGET + find $(TARGET_DIR)/usr/share/heartbeat -name '*.py[co]' -print0 | xargs -0 rm -f +endef +endif + +TARGET_FINALIZE_HOOKS += HEARTBEAT_FINALIZE_TARGET + +$(eval $(autotools-package))
package/libqb/Config.in new file +11 -0
@@ -0,0 +1,11 @@ +config BR2_PACKAGE_LIBQB + bool "libqb" + help + libqb is a library with the primary purpose of providing high + performance client server reusable features. It provides high + performance logging, tracing, ipc, and poll. + + The initial features of libqb comes from the features of corosync that + were thought to useful to other projects. + + https://github.com/ClusterLabs/libqb/wiki
package/libqb/libqb.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 11d69568a3b99c841535fdeacdad92f0d95af73a73beb77a3912dd29dc434744 libqb-0.17.2.tar.gz
package/libqb/libqb.mk new file +20 -0
@@ -0,0 +1,20 @@ +################################################################################ +# +# libqb +# +################################################################################ + +LIBQB_VERSION = 0.17.2 +LIBQB_SITE = $(call github,ClusterLabs,libqb,v$(LIBQB_VERSION)) +LIBQB_SOURCE = libqb-$(LIBQB_VERSION).tar.gz +LIBQB_LICENSE = LGPLv2.1+ +LIBQB_LICENSE_FILES = COPYING +LIBQB_INSTALL_STAGING = YES +LIBQB_AUTORECONF = YES + +define LIBQB_CREATE_M4_DIR + mkdir -p $(@D)/m4 +endef +LIBQB_POST_PATCH_HOOKS += LIBQB_CREATE_M4_DIR + +$(eval $(autotools-package))
package/pacemaker/Config.in new file +9 -0
@@ -0,0 +1,9 @@ +config BR2_PACKAGE_PACEMAKER + bool "pacemaker" + select BR2_PACKAGE_UTIL_LINUX + select BR2_PACKAGE_UTIL_LINUX_LIBUUID + help + High-Availability cluster resource manager for Heartbeat as + well as Corosync. + + http://clusterlabs.org/
package/pacemaker/pacemaker.hash new file +2 -0
@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 2f73043a5de5e66a47462d37cd68c4c6861f71a03c0ea02eb4ae2b1a550cd7b0 pacemaker-Pacemaker-1.1.14.tar.gz
package/pacemaker/pacemaker.mk new file +32 -0
@@ -0,0 +1,32 @@ +################################################################################ +# +# pacemaker +# +################################################################################ + +PACEMAKER_VERSION = Pacemaker-1.1.14 +PACEMAKER_SITE = $(call github,ClusterLabs,pacemaker,$(PACEMAKER_VERSION)) +PACEMAKER_LICENSE = LGPLv2.1 (library), GPLv2 (programs) +PACEMAKER_LICENSE_FILES = COPYING COPYING.LIB +PACEMAKER_DEPENDENCIES += host-pkgconf host-python libtool util-linux bzip2 libxml2 libxslt corosync heartbeat libqb +PACEMAKER_INSTALL_STAGING = YES +PACEMAKER_AUTORECONF = YES + +PACEMAKER_CONF_ENV += PYTHON=$(HOST_DIR)/usr/bin/python + +# This program configure script expects the proper -I flags in CFLAGS and not +# CPPFLAGS +PACEMAKER_CONF_OPTS += \ + CFLAGS="$(TARGET_CFLAGS) -I$(STAGING_DIR)/usr/include -I$(STAGING_DIR)/usr/include/heartbeat" \ + ac_cv_path_HTML2TXT="" \ + ac_cv_path_HELP2MAN="" \ + ac_cv_path_POD2MAN="" \ + ac_cv_path_ASCIIDOC="" \ + ac_cv_path_PUBLICAN="" \ + ac_cv_path_INKSCAPE="" \ + ac_cv_path_XSLTPROC="" \ + ac_cv_path_XMLCATALOG="" \ + ac_cv_path_FOP="" \ + ac_cv_path_XML2CONFIG="$(STAGING_DIR)/usr/bin/xml2-config" + +$(eval $(autotools-package))
Signature · 1 line
2.7.4
PATCH 23/34 pkg-download: allow git+ssh URLs
In reply to Gilles Chanteperdrix--- package/pkg-download.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
package/pkg-download.mk +1 -1
@@ -228,7 +228,7 @@ define DOWNLOAD_INNER scheme="$($(PKG)_SITE_METHOD)" ; \ fi ; \ case "$$scheme" in \ - git) $($(3)_GIT) && exit ;; \ + git|git+ssh) $($(3)_GIT) && exit ;; \ svn) $($(3)_SVN) && exit ;; \ cvs) $($(3)_CVS) && exit ;; \ bzr) $($(3)_BZR) && exit ;; \
Signature · 1 line
2.7.4
PATCH 24/34 grub2: really disable liblzma
In reply to Gilles Chanteperdrix--- boot/grub2/grub2.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
boot/grub2/grub2.mk +1 -1
@@ -62,7 +62,7 @@ GRUB2_CONF_OPTS = \ --with-platform=$(GRUB2_PLATFORM) \ --disable-grub-mkfont \ --enable-efiemu=no \ - --enable-liblzma=no \ + ac_cv_lib_lzma_lzma_code=no \ --enable-device-mapper=no \ --enable-libzfs=no \ --disable-werror
Signature · 1 line
2.7.4
PATCH 25/34 grub2: allow full install (for installer images)
In reply to Gilles Chanteperdrix--- boot/grub2/Config.in | 10 ++++++++++ boot/grub2/grub2.mk | 2 ++ 2 files changed, 12 insertions(+)
boot/grub2/Config.in +10 -0
@@ -139,6 +139,14 @@ config BR2_TARGET_GRUB2_X86_64_EFI endchoice +config BR2_TARGET_FULL_INSTALL + bool "full install on target" + help + Select this option to install grub on target rather than + generating images. + +if !BR2_TARGET_FULL_INSTALL + if BR2_TARGET_GRUB2_I386_PC config BR2_TARGET_GRUB2_BOOT_PARTITION
@@ -166,4 +174,6 @@ config BR2_TARGET_GRUB2_BUILTIN_CONFIG device and other configuration parameters, but however menu entries cannot be described in this embedded configuration. +endif + endif # BR2_TARGET_GRUB2
boot/grub2/grub2.mk +2 -0
@@ -67,6 +67,7 @@ GRUB2_CONF_OPTS = \ --enable-libzfs=no \ --disable-werror +ifneq ($(BR2_TARGET_FULL_INSTALL),y) # We don't want all the native tools and Grub2 modules to be installed # in the target. So we in fact install everything into the host # directory, and the image generation process (below) will use the
@@ -103,5 +104,6 @@ define GRUB2_EFI_STARTUP_NSH endef GRUB2_POST_INSTALL_TARGET_HOOKS += GRUB2_EFI_STARTUP_NSH endif +endif $(eval $(autotools-package))
Signature · 1 line
2.7.4
PATCH 26/34 reproducibility: cpio: avoid embedding file dates in archive
In reply to Gilles Chanteperdrix--- fs/cpio/cpio.mk | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
fs/cpio/cpio.mk +11 -1
@@ -27,8 +27,18 @@ endif # BR2_ROOTFS_DEVICE_CREATION_STATIC ROOTFS_CPIO_PRE_GEN_HOOKS += ROOTFS_CPIO_ADD_INIT +ifeq ($(BR2_REPRODUCIBLE),y) +define ROOTFS_CPIO_REPRODUCIBLE_TOUCH + find $(TARGET_DIR) -print0 | xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) +endef +else +define ROOTFS_CPIO_REPRODUCIBLE_TOUCH + true +endef +endif + define ROOTFS_CPIO_CMD - cd $(TARGET_DIR) && find . | cpio --quiet -o -H newc > $@ + $(ROOTFS_CPIO_REPRODUCIBLE_TOUCH) && cd $(TARGET_DIR) && find . | cpio --quiet -o -H newc > $@ endef $(BINARIES_DIR)/rootfs.cpio.uboot: $(BINARIES_DIR)/rootfs.cpio host-uboot-tools
Signature · 1 line
2.7.4
PATCH 27/34 package/cpio: allow generating host-cpio
In reply to Gilles ChanteperdrixIn order to be sure to have a cpio binar on host which supports --reproducible.
--- package/cpio/cpio.mk | 6 ++++++ 1 file changed, 6 insertions(+)
package/cpio/cpio.mk +6 -0
@@ -18,3 +18,9 @@ CPIO_DEPENDENCIES += argp-standalone endif $(eval $(autotools-package)) + +HOST_CPIO_DEPENDENCIES = + +$(eval $(host-autotools-package)) + +CPIO_REPRODUCIBLE = $(HOST_DIR)/usr/bin/cpio --reproducible
Signature · 1 line
2.7.4
PATCH 28/34 reproducibility/fs/cpio: generate archive with host-cpio
In reply to Gilles ChanteperdrixUsing the --reproducible option to generate a reproducible archive.
--- fs/cpio/cpio.mk | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
fs/cpio/cpio.mk +8 -8
@@ -27,19 +27,19 @@ endif # BR2_ROOTFS_DEVICE_CREATION_STATIC ROOTFS_CPIO_PRE_GEN_HOOKS += ROOTFS_CPIO_ADD_INIT -ifeq ($(BR2_REPRODUCIBLE),y) -define ROOTFS_CPIO_REPRODUCIBLE_TOUCH - find $(TARGET_DIR) -print0 | xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) +ifneq ($(BR2_REPRODUCIBLE),y) +define ROOTFS_CPIO_CMD + cd $(TARGET_DIR) && find . | cpio --quiet -o -H newc > $@ endef else -define ROOTFS_CPIO_REPRODUCIBLE_TOUCH - true -endef -endif +ROOTFS_CPIO_DEPENDENCIES += host-cpio define ROOTFS_CPIO_CMD - $(ROOTFS_CPIO_REPRODUCIBLE_TOUCH) && cd $(TARGET_DIR) && find . | cpio --quiet -o -H newc > $@ + find $(TARGET_DIR) -print0 | \ + xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) && \ + cd $(TARGET_DIR) && find . | $(CPIO_REPRODUCIBLE) --quiet -o -H newc > $@ endef +endif $(BINARIES_DIR)/rootfs.cpio.uboot: $(BINARIES_DIR)/rootfs.cpio host-uboot-tools $(MKIMAGE) -A $(MKIMAGE_ARCH) -T ramdisk \
Signature · 1 line
2.7.4
PATCH 29/34 cdrkit: add patch allowing to set creation date
In reply to Gilles ChanteperdrixThe patch was posted here:
https://lists.gnu.org/archive/html/bug-cpio/2014-08/msg00000.html
And allows generating reproducible iso images with genisoimage.
--- ...-reproducible-allow-setting-creation-date.patch | 140 +++++++++++++++++++++ 1 file changed, 140 insertions(+) create mode 100644 package/cdrkit/0003-reproducible-allow-setting-creation-date.patch
package/cdrkit/0003-reproducible-allow-setting-creation-date.patch new file +140 -0
@@ -0,0 +1,140 @@ +diff --git a/genisoimage/.eltorito.c.swp b/genisoimage/.eltorito.c.swp +new file mode 100644 +index 0000000..a49fcbd +Binary files /dev/null and b/genisoimage/.eltorito.c.swp differ +diff --git a/genisoimage/genisoimage.1 b/genisoimage/genisoimage.1 +index d05b24a..d69a1d4 100644 +--- a/genisoimage/genisoimage.1 ++++ b/genisoimage/genisoimage.1 +@@ -976,6 +976,12 @@ in the + .I .genisoimagerc + file. + .TP ++.BI \-creation-date " epoch" ++Specifies the date to be used as creation, modification and effective ++date in the volume descriptor and for files and relocations created ++on the fly. Specified as a number of second since ++1970-01-01 00:00:00 +0000 (UTC); if 0, the current time is used. ++.TP + .B \-print\-size + Print estimated filesystem size in multiples of the sector size (2048 bytes) + and exit. This option is needed for +diff --git a/genisoimage/genisoimage.c b/genisoimage/genisoimage.c +index cfd079a..58397e9 100644 +--- a/genisoimage/genisoimage.c ++++ b/genisoimage/genisoimage.c +@@ -169,6 +169,7 @@ char *abstract = ABSTRACT_DEFAULT; + char *volset_id = VOLSET_ID_DEFAULT; + char *volume_id = VOLUME_ID_DEFAULT; + char *system_id = SYSTEM_ID_DEFAULT; ++time_t creation_date = 0; + char *boot_catalog = BOOT_CATALOG_DEFAULT; + char *boot_image = BOOT_IMAGE_DEFAULT; + char *genboot_image = BOOT_IMAGE_DEFAULT; +@@ -405,6 +406,8 @@ struct ld_option { + #define OPTION_ALLOW_LEADING_DOTS 1070 + #define OPTION_PUBLISHER 1071 + ++#define OPTION_CREATION_DATE 1072 ++ + #ifdef JIGDO_TEMPLATE + #define OPTION_JTT_OUTPUT 1101 + #define OPTION_JTJ_OUTPUT 1102 +@@ -522,6 +525,8 @@ static const struct ld_option ld_options[] = + '\0', "FILE", "Check all ISO9660 names from previous session", ONE_DASH}, + {{"copyright", required_argument, NULL, OPTION_COPYRIGHT}, + '\0', "FILE", "Set Copyright filename", ONE_DASH}, ++ {{"creation-date", required_argument, NULL, OPTION_CREATION_DATE}, ++ '\0', NULL, "Set volume creation date", ONE_DASH}, + {{"debug", no_argument, NULL, OPTION_DEBUG}, + '\0', NULL, "Set debug flag", ONE_DASH}, + {{"eltorito-boot", required_argument, NULL, 'b'}, +@@ -1721,6 +1726,22 @@ int main(int argc, char *argv[]) + #endif + } + break; ++ case OPTION_CREATION_DATE: ++ { ++ char *end = 0; ++ ++ creation_date = strtol(optarg, &end, 10); ++ if (!end || *end != 0) { ++#ifdef USE_LIBSCHILY ++ comerrno(EX_BAD, "Bad epoch for -creation-date\n"); ++#else ++ fprintf(stderr, "Bad epoch for -creation-date\n"); ++ exit(1); ++#endif ++ } ++ break; ++ } ++ + case OPTION_DEBUG: + debug++; + break; +diff --git a/genisoimage/genisoimage.h b/genisoimage/genisoimage.h +index bbedfb0..c49576c 100644 +--- a/genisoimage/genisoimage.h ++++ b/genisoimage/genisoimage.h +@@ -650,6 +650,7 @@ extern char *appid; + extern char *volset_id; + extern char *system_id; + extern char *volume_id; ++extern time_t creation_date; + extern char *boot_catalog; + extern char *boot_image; + extern char *genboot_image; +diff --git a/genisoimage/tree.c b/genisoimage/tree.c +index 7805888..f17a662 100644 +--- a/genisoimage/tree.c ++++ b/genisoimage/tree.c +@@ -783,7 +783,11 @@ generate_reloc_directory() + struct directory_entry *s_entry; + + /* Create an entry for our internal tree */ +- time(¤t_time); ++ if (creation_date == 0) { ++ time(¤t_time); ++ } else { ++ current_time = creation_date; ++ } + reloc_dir = (struct directory *) + e_malloc(sizeof (struct directory)); + memset(reloc_dir, 0, sizeof (struct directory)); +@@ -2680,7 +2684,11 @@ init_fstatbuf() + time_t current_time; + + if (fstatbuf.st_ctime == 0) { +- time(¤t_time); ++ if (creation_date == 0) { ++ time(¤t_time); ++ } else { ++ current_time = creation_date; ++ } + if (rationalize_uid) + fstatbuf.st_uid = uid_to_use; + else +diff --git a/genisoimage/write.c b/genisoimage/write.c +index a423ab1..f63507c 100644 +--- a/genisoimage/write.c ++++ b/genisoimage/write.c +@@ -1885,12 +1885,17 @@ pvd_write(FILE *outfile) + int should_write; + struct tm local; + struct tm gmt; ++ time_t pvd_date; + + + time(&begun); + +- local = *localtime(&begun); +- gmt = *gmtime(&begun); ++ if (creation_date == 0) { ++ creation_date = begun; ++ } ++ ++ local = *localtime(&creation_date); ++ gmt = *gmtime(&creation_date); + + /* + * There was a comment here about breaking in the year 2000.
Signature · 1 line
2.7.4
PATCH 30/34 reproducibility/fs/iso9660: reproducible iso images
In reply to Gilles ChanteperdrixWhen reproducibility is required, we:
- fix the iso files timestamps to avoid depending on the build date
- run genisoimage with the -creation-date option to avoid using the build date for files in the iso image automatically created by genisoimage (such as the catalog)
- disable the rockridge extension, as it stores each file "ctime", which can not be changed easily on the host filesystem
- force using an initrd, as a trying and booting an iso image without the rockridge extensions will not work
- use isohybrid -i option to avoid embedding a random number in the generated iso.
--- fs/iso9660/Config.in | 9 +++++++++ fs/iso9660/iso9660.mk | 22 ++++++++++++++++++++-- 2 files changed, 29 insertions(+), 2 deletions(-)
fs/iso9660/Config.in +9 -0
@@ -71,6 +71,15 @@ config BR2_TARGET_ROOTFS_ISO9660_INITRD contain a kernel image, an initrd image (unless an initramfs linked into the kernel is used) and the bootloader. +config BR2_TARGET_ROOTFS_ISO9660_REPRODUCIBLE + def_bool BR2_REPRODUCIBLE + select BR2_TARGET_ROOTFS_ISO9660_INITRD + +if BR2_REPRODUCIBLE +comment "A reproducible build requires disabling the RockRidge extensions" +comment "Which makes using an initrd the only sane solution" +endif + config BR2_TARGET_ROOTFS_ISO9660_HYBRID bool "Build hybrid image" depends on BR2_TARGET_ROOTFS_ISO9660_ISOLINUX
fs/iso9660/iso9660.mk +20 -2
@@ -143,15 +143,33 @@ ROOTFS_ISO9660_PRE_GEN_HOOKS += ROOTFS_ISO9660_DISABLE_EXTERNAL_INITRD endif # ROOTFS_ISO9660_USE_INITRD +ifneq ($(BR2_REPRODUCIBLE),y) +define ROOTFS_ISO9660_GENISOIMAGE + $(HOST_DIR)/usr/bin/genisoimage -R +endef +define ROOTFS_ISO9660_ISOHYBRID + $(HOST_DIR)/usr/bin/isohybrid +endef +else +define ROOTFS_ISO9660_GENISOIMAGE + find $(ROOTFS_ISO9660_TARGET_DIR) -print0 | \ + xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) && \ + $(HOST_DIR)/usr/bin/genisoimage -creation-date $(SOURCE_DATE_EPOCH) +endef +define ROOTFS_ISO9660_ISOHYBRID + $(HOST_DIR)/usr/bin/isohybrid -i $(SOURCE_DATE_EPOCH) +endef +endif + define ROOTFS_ISO9660_CMD - $(HOST_DIR)/usr/bin/genisoimage -J -R -b $(ROOTFS_ISO9660_BOOT_IMAGE) \ + $(ROOTFS_ISO9660_GENISOIMAGE) -J -b $(ROOTFS_ISO9660_BOOT_IMAGE) \ -no-emul-boot -boot-load-size 4 -boot-info-table \ -o $@ $(ROOTFS_ISO9660_TARGET_DIR) endef ifeq ($(BR2_TARGET_ROOTFS_ISO9660_HYBRID),y) define ROOTFS_ISO9660_GEN_HYBRID - $(HOST_DIR)/usr/bin/isohybrid -t 0x96 $@ + $(ROOTFS_ISO9660_ISOHYBRID) -t 0x96 $@ endef ROOTFS_ISO9660_POST_GEN_HOOKS += ROOTFS_ISO9660_GEN_HYBRID
Signature · 1 line
2.7.4
PATCH 31/34 reproducible/syslinux: make syslinux build reproducible
In reply to Gilles ChanteperdrixBuild with the target toolchain so that the binaries are identical with different host toolchains.
Sort files lists in order to get deterministic link order. Build with HEXDATE set to the source date epoch.
--- boot/syslinux/0001-fixed-build-order.patch | 37 +++++++++++++++++++++++ boot/syslinux/syslinux.mk | 47 +++++++++++++++++++++++++----- fs/iso9660/iso9660.mk | 4 +-- 3 files changed, 79 insertions(+), 9 deletions(-) create mode 100644 boot/syslinux/0001-fixed-build-order.patch
boot/syslinux/0001-fixed-build-order.patch new file +37 -0
@@ -0,0 +1,37 @@ +--- syslinux-6.03/core/Makefile~ 2014-10-06 18:27:44.000000000 +0200 ++++ syslinux-6.03/core/Makefile 2016-03-24 14:36:38.207391899 +0100 +@@ -41,9 +41,9 @@ + # All primary source files for the main syslinux files + NASMSRC := $(wildcard $(SRC)/*.asm) + NASMHDR := $(wildcard $(SRC)/*.inc) +-CSRC := $(shell find $(SRC) -name '*.c' -print) +-SSRC := $(shell find $(SRC) -name '*.S' -print) +-CHDR := $(shell find $(SRC) -name '*.h' -print) ++CSRC := $(shell find $(SRC) -name '*.c' -print | sort) ++SSRC := $(shell find $(SRC) -name '*.S' -print | sort) ++CHDR := $(shell find $(SRC) -name '*.h' -print | sort) + OTHERSRC := keywords + ALLSRC = $(NASMSRC) $(NASMHDR) $(CSRC) $(SSRC) $(CHDR) $(OTHERSRC) + +@@ -56,18 +56,18 @@ + $(addprefix $(SRC)/fs/pxe/, dhcp_option.c pxe.c tftp.c urlparse.c bios.c) + + LPXELINUX_CSRC = $(CORE_PXE_CSRC) \ +- $(shell find $(SRC)/lwip -name '*.c' -print) \ ++ $(shell find $(SRC)/lwip -name '*.c' -print | sort) \ + $(addprefix $(SRC)/fs/pxe/, \ + core.c dnsresolv.c ftp.c ftp_readdir.c gpxeurl.c http.c \ + http_readdir.c idle.c isr.c tcp.c) + + PXELINUX_CSRC = $(CORE_PXE_CSRC) \ +- $(shell find $(SRC)/legacynet -name '*.c' -print) ++ $(shell find $(SRC)/legacynet -name '*.c' -print | sort) + + LPXELINUX_OBJS = $(subst $(SRC)/,,$(LPXELINUX_CSRC:%.c=%.o)) + PXELINUX_OBJS = $(subst $(SRC)/,,$(PXELINUX_CSRC:%.c=%.o)) + +-UNITTEST_CSRC = $(shell find $(SRC) -path '*/tests/*.c' -print) ++UNITTEST_CSRC = $(shell find $(SRC) -path '*/tests/*.c' -print | sort) + UNITTEST_OBJS = $(subst $(SRC)/,,$(UNITTEST_CSRC:%.c=%.o)) + + # Don't include console and network stack specific objects or unit tests
boot/syslinux/syslinux.mk +40 -7
77 lines not on this page · raw message
fs/iso9660/iso9660.mk +2 -2
@@ -70,8 +70,6 @@ ROOTFS_ISO9660_BOOT_IMAGE = isolinux/isolinux.bin define ROOTFS_ISO9660_INSTALL_BOOTLOADER $(INSTALL) -D -m 0644 $(BINARIES_DIR)/syslinux/* \ $(ROOTFS_ISO9660_TARGET_DIR)/isolinux/ - $(INSTALL) -D -m 0644 $(HOST_DIR)/usr/share/syslinux/ldlinux.c32 \ - $(ROOTFS_ISO9660_TARGET_DIR)/isolinux/ldlinux.c32 endef endif
@@ -168,6 +166,8 @@ define ROOTFS_ISO9660_CMD endef ifeq ($(BR2_TARGET_ROOTFS_ISO9660_HYBRID),y) +ROOTFS_ISO9660_DEPENDENCIES += host-syslinux + define ROOTFS_ISO9660_GEN_HYBRID $(ROOTFS_ISO9660_ISOHYBRID) -t 0x96 $@ endef
Signature · 1 line
2.7.4
PATCH 32/34 reproducible/iso9660: generate reproducible isos
In reply to Gilles ChanteperdrixSort the files list to get a deterministic file order
--- fs/iso9660/iso9660.mk | 25 +++++++++++++------------ 1 file changed, 13 insertions(+), 12 deletions(-)
fs/iso9660/iso9660.mk +13 -12
@@ -142,29 +142,30 @@ endif # ROOTFS_ISO9660_USE_INITRD ifneq ($(BR2_REPRODUCIBLE),y) -define ROOTFS_ISO9660_GENISOIMAGE - $(HOST_DIR)/usr/bin/genisoimage -R +define ROOTFS_ISO9660_CMD + $(HOST_DIR)/usr/bin/genisoimage -R -J -b $(ROOTFS_ISO9660_BOOT_IMAGE) \ + -no-emul-boot -boot-load-size 4 -boot-info-table \ + -o $@ $(ROOTFS_ISO9660_TARGET_DIR) endef define ROOTFS_ISO9660_ISOHYBRID $(HOST_DIR)/usr/bin/isohybrid endef else -define ROOTFS_ISO9660_GENISOIMAGE - find $(ROOTFS_ISO9660_TARGET_DIR) -print0 | \ - xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) && \ - $(HOST_DIR)/usr/bin/genisoimage -creation-date $(SOURCE_DATE_EPOCH) +define ROOTFS_ISO9660_CMD + find $(ROOTFS_ISO9660_TARGET_DIR) -print | \ + xargs touch -h -d @$(SOURCE_DATE_EPOCH) && \ + find $(ROOTFS_ISO9660_TARGET_DIR) -! -type d -o -empty | sort | \ + sed "s%$(ROOTFS_ISO9660_TARGET_DIR)"'\\(/.*\\)%\\1=&%' | \ + $(HOST_DIR)/usr/bin/genisoimage -J -creation-date $(SOURCE_DATE_EPOCH) \ + -b $(ROOTFS_ISO9660_BOOT_IMAGE) \ + -no-emul-boot -boot-load-size 4 -boot-info-table \ + -o $@ -graft-points -path-list - endef define ROOTFS_ISO9660_ISOHYBRID $(HOST_DIR)/usr/bin/isohybrid -i $(SOURCE_DATE_EPOCH) endef endif -define ROOTFS_ISO9660_CMD - $(ROOTFS_ISO9660_GENISOIMAGE) -J -b $(ROOTFS_ISO9660_BOOT_IMAGE) \ - -no-emul-boot -boot-load-size 4 -boot-info-table \ - -o $@ $(ROOTFS_ISO9660_TARGET_DIR) -endef - ifeq ($(BR2_TARGET_ROOTFS_ISO9660_HYBRID),y) ROOTFS_ISO9660_DEPENDENCIES += host-syslinux
Signature · 1 line
2.7.4
PATCH 33/34 reproducible/cpio: generate reproducible cpio archives
In reply to Gilles ChanteperdrixSort the file list to get a deterministic file order.
--- fs/cpio/cpio.mk | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
fs/cpio/cpio.mk +3 -3
@@ -35,9 +35,9 @@ else ROOTFS_CPIO_DEPENDENCIES += host-cpio define ROOTFS_CPIO_CMD - find $(TARGET_DIR) -print0 | \ - xargs -0 touch -h -d @$(SOURCE_DATE_EPOCH) && \ - cd $(TARGET_DIR) && find . | $(CPIO_REPRODUCIBLE) --quiet -o -H newc > $@ + find $(TARGET_DIR) | \ + xargs touch -h -d @$(SOURCE_DATE_EPOCH) && \ + cd $(TARGET_DIR) && find . | sort | $(CPIO_REPRODUCIBLE) --quiet -o -H newc > $@ endef endif
Signature · 1 line
2.7.4
PATCH 34/34 reproducible/grub2: get a reproducible grub2 full installation
In reply to Gilles ChanteperdrixGet a reproducible modules dependencies file by fixing the awk script to avoid depending on awk hash order, in order to get the same result independently of the particular variant of awk used.
Compile grub2 with the target toolchain in the "full install" case, so that the generated binaries do not depend on the host toolchain used.
--- boot/grub2/grub2-deterministic-awk.patch | 68 ++++++++++++++++++++++++++++++++ boot/grub2/grub2.mk | 34 ++++++++++------ 2 files changed, 91 insertions(+), 11 deletions(-) create mode 100644 boot/grub2/grub2-deterministic-awk.patch
boot/grub2/grub2-deterministic-awk.patch new file +68 -0
@@ -0,0 +1,68 @@ +--- grub2-2.00/grub-core/genmoddep.awk~ 2016-03-27 15:36:25.138050578 +0200 ++++ grub2-2.00/grub-core/genmoddep.awk 2016-03-27 15:39:33.704636947 +0200 +@@ -15,17 +15,21 @@ + BEGIN { + error = 0 + lineno = 0; ++ i = 1; + while (getline <"/dev/stdin") { + lineno++; + if ($1 == "defined") { + symtab[$3] = $2; +- modtab[$2] = "" modtab[$2] ++ if (!modtab_index[$2]) ++ modtab_index[$2]=i++ + } else if ($1 == "undefined") { ++ if (!modtab_index[$2]) ++ modtab_index[$2]=i++ + if ($3 in symtab) +- modtab[$2] = modtab[$2] " " symtab[$3]; ++ modtab[modtab_index[$2]] = modtab[modtab_index[$2]] " " symtab[$3]; + else if ($3 != "__gnu_local_gp") { +- printf "%s in %s is not defined\n", $3, $2 >"/dev/stderr"; +- error++; ++ printf "%s in %s is not defined\n", $3, $2 >"/dev/stderr"; ++ error++; + } + } + else { +@@ -41,24 +45,28 @@ END { + if (error >= 1) + exit 1; + +- for (mod in modtab) { ++ for (j = 1; j < i; j++) { ++ for (mod in modtab_index) ++ if (modtab_index[mod] == j) ++ break; + # Remove duplications. +- split(modtab[mod], depmods, " "); ++ len=split(modtab[j], depmods, " "); + for (depmod in uniqmods) { + delete uniqmods[depmod]; + } +- for (i in depmods) { +- depmod = depmods[i]; +- # Ignore kernel, as always loaded. +- if (depmod != "kernel" && depmod != mod) +- uniqmods[depmod] = 1; +- } + modlist = "" +- for (depmod in uniqmods) { +- modlist = modlist " " depmod; ++ for (k = 1; k <= len; k++) { ++ depmod = depmods[k] ++ if (uniqmods[depmod]) ++ continue ++ # Ignore kernel, as always loaded. ++ if (depmod != "kernel" && depmod != mod) { ++ uniqmods[depmod] = 1; ++ modlist = modlist " " depmod; ++ } + } + if (mod == "all_video") { +- continue; ++ continue; + } + printf "%s:%s\n", mod, modlist; + }
boot/grub2/grub2.mk +23 -11
@@ -46,17 +46,6 @@ endif # bootloader itself; none of these are used to build the native # tools. -GRUB2_CONF_ENV = \ - $(HOST_CONFIGURE_OPTS) \ - CPP="$(HOSTCC) -E" \ - TARGET_CC="$(TARGET_CC)" \ - TARGET_CFLAGS="$(TARGET_CFLAGS)" \ - TARGET_CPPFLAGS="$(TARGET_CPPFLAGS)" \ - TARGET_LDFLAGS="$(TARGET_LDFLAGS)" \ - NM="$(TARGET_NM)" \ - OBJCOPY="$(TARGET_OBJCOPY)" \ - STRIP="$(TARGET_STRIP)" - GRUB2_CONF_OPTS = \ --target=$(GRUB2_TARGET) \ --with-platform=$(GRUB2_PLATFORM) \
@@ -68,6 +57,17 @@ GRUB2_CONF_OPTS = \ --disable-werror ifneq ($(BR2_TARGET_FULL_INSTALL),y) +GRUB2_CONF_ENV = \ + $(HOST_CONFIGURE_OPTS) \ + CPP="$(HOSTCC) -E" \ + TARGET_CC="$(TARGET_CC)" \ + TARGET_CFLAGS="$(TARGET_CFLAGS)" \ + TARGET_CPPFLAGS="$(TARGET_CPPFLAGS)" \ + TARGET_LDFLAGS="$(TARGET_LDFLAGS)" \ + NM="$(TARGET_NM)" \ + OBJCOPY="$(TARGET_OBJCOPY)" \ + STRIP="$(TARGET_STRIP)" + # We don't want all the native tools and Grub2 modules to be installed # in the target. So we in fact install everything into the host # directory, and the image generation process (below) will use the
@@ -104,6 +104,18 @@ define GRUB2_EFI_STARTUP_NSH endef GRUB2_POST_INSTALL_TARGET_HOOKS += GRUB2_EFI_STARTUP_NSH endif + +else +GRUB2_CONF_ENV = \ + $(TARGET_CONFIGURE_OPTS) \ + CPP="$(TARGET_CC) -E" \ + TARGET_CC="$(TARGET_CC)" \ + TARGET_CFLAGS="$(TARGET_CFLAGS)" \ + TARGET_CPPFLAGS="$(TARGET_CPPFLAGS)" \ + TARGET_LDFLAGS="$(TARGET_LDFLAGS)" \ + NM="$(TARGET_NM)" \ + OBJCOPY="$(TARGET_OBJCOPY)" \ + STRIP="$(TARGET_STRIP)" endif $(eval $(autotools-package))
Signature · 1 line
2.7.4
Hello Gilles,
Quoted 7 lines · Gilles Chanteperdrix
some time ago, I worked on getting reproducible builds for buildroot. I take "reproducible builds" in a broader sense than what Debian does for instance: in the case of buildroot, we would like the built binaries to be as independent of the build environment as possible, so that in addition to getting identical binaries when building twice on the same machine, we also get identical binaries when building on a different distribution.
First of all, thanks a lot for working on this! It's definitely an interesting and useful topic.
Quoted 11 lines
So far, I got
reproducible builds for a (relatively reduced) subset of buildroot packages, on a Debian jessie and Slackware 14.1, with as many differences in the distributions configuration as I could find (timezone, locale, date, time, etc... even the awk flavor used). The only thing I have to get identical is the build path, this is what Debian does, a patch has been proposed a long time ago to allow gcc to avoid the dependence between the__FILE__macro and the build path, but it does not seem to have been merged:
https://mail-index.netbsd.org/tech-toolchain/2009/02/17/msg000577.html
Quoted 7 lines
I expected to be able to clean up the patches before submission (as many patches do the same thing, like overriding the
__DATE__and__TIME__macros in order to avoid depending on the build time, so this could probably be made generic by buildroot core), but I am not going to have time soon to do that, so here come the patches as they are, in order to spark discussion and gather remarks, before I have time to do more. The patches are based on the 2015.11.1 release.
Indeed, as you say yourself, some of your patches are not mergeable as-is. However, several of them are not related, or directly related to reproducible builds (some of your patches add new packages, etc.).
One first thing that is missing is your Signed-off-by line on the patches, which we require for all Buildroot contributions (exactly like for the Linux kernel).
Another thing that worries me is all the patches/modifications needed by each package to make their build "reproducible". Do you expect such patches to be merged in the respective upstream projects?
Finally, I'm also concerned by the testability of the reproducible build feature. To make sure such a feature work, we would need to do some automated build testing of the same configuration in various environments/configurations and ensure they provide the same output. This is clearly not impossible to do, but there's quite some work involved to set up such an infrastructure.
I'll review some of your patches (when it makes sense). However, I'll mark them all as "RFC" in our patch tracking system: since none of them have a Signed-off-by line, we can't even take them. It would be good if you could resend first with your SoB line the patches that are not related to the reproducible builds. It would help reduce your stack of patches.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 1/34 reproducibility: introduce config knob
In reply to Gilles ChanteperdrixHello,
Quoted 3 lines · Gilles Chanteperdrix
Making the builds reproducible involves turning off build timestamps, which some users may find annoying. So make the reproducible builds optional.
Missing SoB, as I said.
Originally, I thought "why make this optional", but I indeed understand that for some use cases, not having timestamps might be annoying. So I guess making this optional is the best option.
Config.in config BR2_COMPILER_PARANOID_UNSAFE_PATH · line 627
Quoted 3 lines
Config.in +2
-0+config BR2_REPRODUCIBLE + bool "Try and remove all sources of non-reproducibility from the build"
This should probably be:
bool "Make the build reproducible (experimental)" help
This option will remove all sources of non-reproducibility from the build process. For a given Buildroot configuration, this allows to generate exactly identical binaries from one build to the other, including on different machines.
This is labeled as an experimental feature, as not all packages behave properly to ensure reproducibility.
Best regards,
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 2/34 reproducibility: override locale and timezone
In reply to Gilles ChanteperdrixHello,
Quoted 3 lines · Gilles Chanteperdrix
--- Makefile | 7 +++++++ 1 file changed, 7 insertions(+)
Missing SoB.
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 221
Quoted 9 lines
Makefile +1
-0@@ -218,6 +218,13 @@ ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) -include $(BR2_CONFIG) endif +export BR2_REPRODUCIBLE
Why ?
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 226
Quoted 5 lines
Makefile +5
-0+ifeq ($(BR2_REPRODUCIBLE),y) +export TZ=UTC +export LANG=C +export LC_ALL=C +endif
Regarding overriding TZ, LANG and LC_ALL, I'm wondering if we shouldn't do it unconditionally, regardless of the value of BR2_REPRODUCIBLE. What is the benefit of leaving those variable to their value from the build environment?
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Gilles ChanteperdrixHello,
Quoted 1 line · Gilles Chanteperdrix
When reproducibility is wanted, generates a global
SOURCE_DATE_EPOCH
wanted -> requested
generates -> generate
Quoted 7 lines
environment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.
This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
Missing SoB.
Makefile ifeq ($(BR2_REPRODUCIBLE),y) · line 229
Quoted 4 lines
Makefile +4
-0+SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at) +SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s)
Yirk. I'm wondering if we shouldn't simply have a BR2_RELEASE_DATE variable next to BR2_VERSION in the top level Makefile that gives the release date.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 4/34 reproducibility: make rootfs.tar reproducible
In reply to Gilles ChanteperdrixHello,
fs/tar/tar.mk line 19
Quoted 6 lines · Gilles Chanteperdrix
fs/tar/tar.mk +6
-0+define ROOTFS_TAR_CMD + cd $(TARGET_DIR) && { \ + find . -\( -! -type d -o -empty -\) -print0 | \ + sort -z | \ + tar --null -T - -c$(TAR_OPTS)f $@ --mtime=@$(SOURCE_DATE_EPOCH) --numeric-owner; \ + }
We normally write such constructs as:
(cd $(TARGET_DIR) && \ foo ....)
However, this raises the question of what's needed for all the other filesystem formats. Will they all have to implement a different ROOTFS_<foo>_CMD variable ? Or will there be some commonalities that should be factored out in the common rootfs image infrastructure ?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 7/34 reproducibility/libgcrypt: override timestamps
In reply to Gilles ChanteperdrixHello,
package/libgcrypt/0003-honor-source-date-epoch.patch
Quoted 12 lines · Gilles Chanteperdrix
with
SOURCE_DATE_EPOCH--- package/libgcrypt/0003-honor-source-date-epoch.patch | 15 +++++++++++++++ package/libgcrypt/libgcrypt.mk | 1 + 2 files changed, 16 insertions(+) create mode 100644 package/libgcrypt/0003-honor-source-date-epoch.patchpackage/libgcrypt/0003-honor-source-date-epoch.patch new file +0
-0
All patches need a description + Signed-off-by.
Also, has this been submitted upstream? Is it already used by Debian?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Gilles ChanteperdrixHello,
Cc'ing Gustavo on this, since he has done quite a bit of libglib packaging work.
Quoted 6 lines · Gilles Chanteperdrix
glib2 installs compiled python bytecode in
/usr/share/glib-2.0/codegen, likely for the gdbus-codegen program.This bytecode is not compiled with the buildroot host python, which causes a first source of non reproducibility, fix that by using buildroot host python.
Which has the annoying side effect of increasing even more the set of mandatory dependencies to build libglib :-/
So I wonder if the dependency on host-python should be made conditional on BR2_REPRODUCIBLE.
Quoted 5 lines
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries, so, since buildroot users are unlikely to need the qdbus-codegen program on target, add an option to remove it. The option is disabled by default.
Since then, Gustavo has already modified the libglib2 package to remove the /usr/share/glib-2.0/codegen directory. However, gdbus-codegen is not removed.
See commit 1f1f16e9e5367048faa1cf17237e0c0d422e98d6.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 11/34 reproducibility/vim: make vim reproducible
In reply to Gilles ChanteperdrixHello,
Quoted 5 lines · Gilles Chanteperdrix
package/vim/vim.mk +5
-0+ifeq ($(BR2_REPRODUCIBLE),y) +VIM_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +VIM_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +VIM_CONF_OPTS += \
package/vim/vim.mk else · line 51
Quoted 10 lines
package/vim/vim.mk +10
-0+ --with-compiledby=buildroot \ + CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(VIM_DIR)/.br_date_override.h" + +define VIM_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(VIM_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(VIM_SOURCE_DATE))" >> $(VIM_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(VIM_SOURCE_TIME))" >> $(VIM_DIR)/.br_date_override.h +endef + +VIM_POST_PATCH_HOOKS += VIM_OVERRIDE_DATE_HOOK
As you say yourself, this seems to be needed in many packages, so I think a more generic solution is needed. At the very least, the .br_date_override.h file should be generated once in a generic place.
package/vim/vim.mk else · line 57
Quoted 5 lines
package/vim/vim.mk +5
-0+define VIM_FIX_AWK_REFERENCE + sed -i 's,^#!.*awk,#!/usr/bin/awk,' $(TARGET_DIR)/usr/share/vim/vim74/tools/mve.awk +endef + +VIM_POST_INSTALL_TARGET_HOOKS += VIM_FIX_AWK_REFERENCE
Do we even need this awk script installed?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 12/34 reproducibility/wireshark: remove documentation
In reply to Gilles ChanteperdrixHello,
Quoted 7 lines · Gilles Chanteperdrix
HTML documentation contains a link to a mail address specific to the distribution of the build machine. Remove html documentation from the rootfs as it is unlikely to be needed. At this chance also remove the txt documentation.
--- package/wireshark/wireshark.mk | 6 ++++++ 1 file changed, 6 insertions(+)
If this one had your Signed-off-by, I could have applied it. Can you respin?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 2/34 reproducibility: override locale and timezone
In reply to Thomas PetazzoniQuoted 17 lines · Thomas Petazzoni
Hello,
Quoted 3 lines · Gilles Chanteperdrix
--- Makefile | 7 +++++++ 1 file changed, 7 insertions(+)Missing SoB.
Quoted 8 lines
Makefile +0
-0@@ -218,6 +218,13 @@ ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) -include $(BR2_CONFIG) endif
Quoted 3 lines
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 221
Quoted 1 line
Makefile +1
-0+export BR2_REPRODUCIBLEWhy ?
I guess because it will (or may be) be used in some helper scripts...
Quoted 11 lines
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 226
Quoted 6 lines
Makefile +5
-0+ifeq ($(BR2_REPRODUCIBLE),y) +export TZ=UTC +export LANG=C +export LC_ALL=C +endifRegarding overriding TZ, LANG and
LC_ALL, I'm wondering if we shouldn't do it unconditionally, regardless of the value ofBR2_REPRODUCIBLE. What is the benefit of leaving those variable to their value from the build environment?
For the record: I had the same comment about this entire series. Most of it
should be done unconditionally, only the aspects that may be annoying for debugging (like the timestamps) should be conditional.
Unfortunately, that will be a cause for discussion... For instance, LANG=C
means that compiler error messages will be in English instead of in your native language. Do we really want that by default?
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 4/34 reproducibility: make rootfs.tar reproducible
In reply to Thomas PetazzoniQuoted 13 lines · Thomas Petazzoni
Hello,
fs/tar/tar.mk line 19
Quoted 6 lines · Gilles Chanteperdrix
fs/tar/tar.mk +6
-0+define ROOTFS_TAR_CMD + cd $(TARGET_DIR) && { \ + find . -\( -! -type d -o -empty -\) -print0 | \ + sort -z | \ + tar --null -T - -c$(TAR_OPTS)f $@ --mtime=@$(SOURCE_DATE_EPOCH) --numeric-owner; \ + }We normally write such constructs as:
Quoted 2 lines
(cd $(TARGET_DIR) && \ foo ....)
Actually, we don't AFAIK... In general, the parenthesis are not needed so they
should be removed. So also in this case it should be
cd $(TARGET_DIR); find ....
Quoted 5 lines
However, this raises the question of what's needed for all the other filesystem formats. Will they all have to implement a different ROOTFS_<foo>_CMD variable ? Or will there be some commonalities that should be factored out in the common rootfs image infrastructure ?
Yes, to me it makes more sense to do this in the actual target directory after
the post-build scripts have been run.
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 13/34 reproducibility/libtool: fix reproducibility issues
In reply to Gilles ChanteperdrixHello,
This probably should be several separate patches.
package/libtool/0001-fix-cross-compilation.patch
Quoted 16 lines · Gilles Chanteperdrix
- remove libtool and libtoolize; they are full of references to the build environment tools, and unlikely to be needed on target;
- do not use/etc/ld.so.confcontents in libtool search path: they are not relevant on target, and cause a useless dependency of libltdl on the build environment.--- package/libtool/0001-fix-cross-compilation.patch | 11 +++++++++++ package/libtool/libtool.mk | 8 ++++++-- 2 files changed, 17 insertions(+), 2 deletions(-) create mode 100644 package/libtool/0001-fix-cross-compilation.patchpackage/libtool/0001-fix-cross-compilation.patch new file +0
-0
This patch should have a description + Signed-off-by.
However, libtool is really a core package, and not a simple one, so I'm a bit uneasy with the idea of carrying patches against it. What are the chances of having something like this submitted upstream?
package/libtool/libtool.mk LIBTOOL_VERSION = 2.4.6 · line 11
Quoted 11 lines
package/libtool/0001-fix-cross-compilation.patch new file +1
-0+ fipackage/libtool/libtool.mk +1
-1@@ -8,7 +8,7 @@ LIBTOOL_VERSION = 2.4.6 LIBTOOL_SOURCE = libtool-$(LIBTOOL_VERSION).tar.xz LIBTOOL_SITE = $(BR2_GNU_MIRROR)/libtool LIBTOOL_INSTALL_STAGING = YES -LIBTOOL_CONF_ENV = HELP2MAN=true +LIBTOOL_CONF_OPTS = HELP2MAN=true
Why?
package/libtool/libtool.mk define LIBTOOL_AVOID_AUTORECONF_HOOK · line 36
Quoted 8 lines
package/libtool/libtool.mk +0
-1LIBTOOL_DEPENDENCIES = host-m4 LIBTOOL_LICENSE = GPLv2+ LIBTOOL_LICENSE_FILES = COPYING@@ -33,9 +33,13 @@ define LIBTOOL_AVOID_AUTORECONF_HOOK find $(@D) -name configure -exec touch '{}' \; find $(@D) -name Makefile.in -exec touch '{}' \; endef -LIBTOOL_PRE_CONFIGURE_HOOKS += LIBTOOL_AVOID_AUTORECONF_HOOK
Why?
package/libtool/libtool.mk define LIBTOOL_AVOID_AUTORECONF_HOOK · line 41
Quoted 6 lines
package/libtool/libtool.mk +4
-0HOST_LIBTOOL_PRE_CONFIGURE_HOOKS += LIBTOOL_AVOID_AUTORECONF_HOOK +define LIBTOOL_REMOVE_LIBTOOL + rm $(TARGET_DIR)/usr/bin/libtool $(TARGET_DIR)/usr/bin/libtoolize +endef +LIBTOOL_POST_INSTALL_TARGET_HOOKS += LIBTOOL_REMOVE_LIBTOOL
This part could be a separate patch that would be easy to apply today, so don't hesitate to resend just this part, with your Signed-off-by line, of course.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 14/34 reproducibility/python: also remove pyo files
In reply to Gilles ChanteperdrixHello,
Quoted 3 lines · Gilles Chanteperdrix
when
BR2_PACKAGE_PYTHON_PY_ONLYis selected. Make that option the default whenBR2_REPRODUCIBLEis selected in order to avoid problems with timestamps embedded in compiled python files.
The title of the patch says "also remove pyo files", but this patch does a lot more.
package/python/Config.in if BR2_PACKAGE_PYTHON · line 23
Quoted 11 lines
package/python/Config.in +2
-1@@ -19,7 +19,8 @@ if BR2_PACKAGE_PYTHON choice prompt "python module format to install" - default BR2_PACKAGE_PYTHON_PYC_ONLY + default BR2_PACKAGE_PYTHON_PYC_ONLY if !BR2_REPRODUCIBLE + default BR2_PACKAGE_PYTHON_PY_ONLY if BR2_REPRODUCIBLE
If .pyc files are really not suitable for reproducible builds, then the option of having .pyc files should be completely hidden when BR2_REPRODUCIBLE=y.
package/python/python.mk endif · line 228
Quoted 10 lines
package/python/python.mk +1
-1@@ -225,7 +225,7 @@ endif ifeq ($(BR2_PACKAGE_PYTHON_PY_ONLY),y) define PYTHON_FINALIZE_TARGET - find $(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR) -name '*.pyc' -print0 | xargs -0 rm -f + find $(TARGET_DIR)/usr/lib/python$(PYTHON_VERSION_MAJOR) -name '*.py[co]' -print0 | xargs -0 rm -f
This part is already handled in another patch series I have sent around Python. See https://patchwork.ozlabs.org/patch/617272/ specifically.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 15/34 reproducibility/gzip: avoid gzip timestamp in rootfs.tar.gz
In reply to Gilles ChanteperdrixHello,
Quoted 9 lines · Gilles Chanteperdrix
--- fs/common.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)fs/common.mk +0
-0
fs/common.mk ROOTFS_$(2)_DEPENDENCIES += host-fakeroot host-makedevs \ · line 54
Quoted 8 lines
fs/common.mk +1
-1@@ -48,7 +48,7 @@ ROOTFS_$(2)_DEPENDENCIES += host-fakeroot host-makedevs \ ifeq ($$(BR2_TARGET_ROOTFS_$(2)_GZIP),y) ROOTFS_$(2)_COMPRESS_EXT = .gz -ROOTFS_$(2)_COMPRESS_CMD = gzip -9 -c +ROOTFS_$(2)_COMPRESS_CMD = gzip -9 -c $(if $(BR2_REPRODUCIBLE),-n) endif ifeq ($$(BR2_TARGET_ROOTFS_$(2)_BZIP2),y) ROOTFS_$(2)_COMPRESS_EXT = .bz2
Looks good, just resend with your Signed-off-by line.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 16/34 reproducibility/lzo: forbid lzo with reproducible builds
In reply to Gilles ChanteperdrixHello,
Quoted 1 line · Gilles Chanteperdrix
The lzo archive contains a timestamp.
Missing Signed-off-by line.
Quoted 17 lines
--- fs/tar/Config.in | 7 +++++++ 1 file changed, 7 insertions(+)fs/tar/Config.in +1
-0@@ -33,6 +33,7 @@ config BR2_TARGET_ROOTFS_TAR_LZMA config BR2_TARGET_ROOTFS_TAR_LZO bool "lzo" + depends on !BR2_REPRODUCIBLE help Do compress the tarball with lzop.
fs/tar/Config.in config BR2_TARGET_ROOTFS_TAR_XZ · line 51
Quoted 8 lines
fs/tar/Config.in +5
-0@@ -43,6 +44,12 @@ config BR2_TARGET_ROOTFS_TAR_XZ endchoice +if BR2_REPRODUCIBLE + +comment "lzo compression is only available with non-reproducible builds" + +endif
When we are in such situations, we normally do:
comment "blablabla"
depends on BR2_REPRODUCIBLE
however, in this specific case, I would simply not bother adding a comment. If we start adding a comment for everything that is not suitable for reproducible builds, we're going to add hundreds of them I believe.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 17/34 reproducibility/libnspr: override build timestamps
In reply to Gilles ChanteperdrixHello,
Quoted 1 line · Gilles Chanteperdrix
with
SOURCE_DATE_EPOCH
Missing SoB line here.
package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch
Quoted 10 lines
--- package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch | 71 ++++++++++++++++++++++ 1 file changed, 71 insertions(+) create mode 100644 package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patchpackage/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch new file +0
-0
Missing patch description and Signed-off-by line. Also, did you submit this upstream? Does Debian carry a similar patch for its Reproducible Builds effort?
Those questions are relevant for every package for which you're adding a patch.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 19/34 reproducibility/python: make python reproducible
In reply to Gilles ChanteperdrixHello,
Quoted 7 lines · Gilles Chanteperdrix
python build is not reproducible for two reasons:
- the libpython library embeds the build timestamps, fix this by overriding the__DATE__and__TIME__macro usingSOURCE_DATE_EPOCH;
-_sysconfigdata.pyand config/Makefile on target includes references through INSTALL andMKDIR_Pvariables to host tools, fix this by overriding these variables values with the corresponding paths on buildroot rootfs.
For this second part, are you really aiming at having reproducible builds with different build directories ? As you say in your cover letter, that's an even more complicated problem.
Quoted 5 lines
package/python/python.mk +5
-0+ifeq ($(BR2_REPRODUCIBLE),y) +PYTHON_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +PYTHON_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +PYTHON_CONF_OPTS += \
package/python/python.mk endif · line 233
Quoted 10 lines
package/python/python.mk +10
-0+ CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(PYTHON_DIR)/.br_date_override.h" + +define PYTHON_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(PYTHON_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(PYTHON_SOURCE_DATE))" >> $(PYTHON_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(PYTHON_SOURCE_TIME))" >> $(PYTHON_DIR)/.br_date_override.h +endef + +PYTHON_POST_PATCH_HOOKS += PYTHON_OVERRIDE_DATE_HOOK +endif
This really needs to be factored out somewhere. Maybe our compiler wrapper should take of this. Arnout?
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 21/34 hddtemp: add
In reply to Gilles ChanteperdrixHello,
Quoted 9 lines · Gilles Chanteperdrix
--- package/Config.in | 1 + package/hddtemp/Config.in | 8 ++++++++ package/hddtemp/hddtemp.hash | 2 ++ package/hddtemp/hddtemp.mk | 19 +++++++++++++++++++ 4 files changed, 30 insertions(+) create mode 100644 package/hddtemp/Config.in create mode 100644 package/hddtemp/hddtemp.hash create mode 100644 package/hddtemp/hddtemp.mk
Generally looks good. As usual your SoB line is missing, so we can't apply for the moment. Some more comments below.
package/hddtemp/Config.in line 2
Quoted 8 lines
package/hddtemp/Config.in new file +2
-0@@ -0,0 +1,8 @@ +config BR2_PACKAGE_HDDTEMP + bool "hddtemp"
Really no toolchain dependencies? Not using fork() ? Not using threads ?
Quoted 7 lines
package/hddtemp/Config.in new file +6
-0+++ b/package/hddtemp/hddtemp.mk@@ -0,0 +1,19 @@ +################################################################################ +# +# hddtemp +# +################################################################################
package/hddtemp/Config.in line 15
Quoted 10 lines
package/hddtemp/Config.in new file +10
-0+ +HDDTEMP_VERSION = 0.3-beta15 +HDDTEMP_SOURCE = hddtemp-$(HDDTEMP_VERSION).tar.bz2 +HDDTEMP_SITE = http://download.savannah.gnu.org/releases/hddtemp +HDDTEMP_LICENSE = GPLv2 +HDDTEMP_LICENSE_FILES = GPL-2 + +define HDDTEMP_REMOVE_MAN + rm -rf $(TARGET_DIR)/usr/man +endef
This is not needed usr/man is removed from $(TARGET_DIR) at the end of the build. See the target-finalize make target in the main Buildroot Makefile.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 22/34 High availability: add packages
In reply to Gilles ChanteperdrixHello,
Quoted 5 lines · Gilles Chanteperdrix
cluster-glue
corosync
heartbeat
libqb
pacemaker
This should be split in a series of patches, with one package per patch.
package/cluster-glue/0001-fix-cross-compile.patch
Quoted 9 lines
menu "Javascript"
package/Config.in +0
-0source "package/angularjs/Config.in" source "package/bootstrap/Config.in"package/cluster-glue/0001-fix-cross-compile.patch new file +0
-0@@ -0,0 +1,26 @@
All patches need a description and Signed-off-by line.
package/cluster-glue/Config.in line 4
Quoted 10 lines
package/cluster-glue/Config.in new file +4
-0@@ -0,0 +1,9 @@ +config BR2_PACKAGE_CLUSTER_GLUE + bool "cluster-glue" + select BR2_PACKAGE_UTIL_LINUX + select BR2_PACKAGE_UTIL_LINUX_LIBUUID
This looks wrong, for two reasons:
(1) Because you are not replicating the dependencies of util-linux here.
(2) Because this package has many more packages in its <pkg>_DEPENDENCIES variable than just util-linux. For example, libxml2, libglib2, etc. They should all be selected here as well.
package/cluster-glue/Config.in
Quoted 6 lines
package/cluster-glue/Config.in new file +6
-0+CLUSTER_GLUE_VERSION = 0a7add1d9996 +CLUSTER_GLUE_SOURCE = $(CLUSTER_GLUE_VERSION).tar.bz2 +CLUSTER_GLUE_SITE = http://hg.linux-ha.org/glue/archive +CLUSTER_GLUE_LICENSE = LGPLv2.1+ (library), GPLv2+ (programs) +CLUSTER_GLUE_LICENSE_FILES = COPYING COPYING.LIB +CLUSTER_GLUE_DEPENDENCIES += host-pkgconf host-python libtool util-linux bzip2 libglib2 libxml2
The dependency on libtool is useless, it's automatically added by <pkg>_AUTORECONF = YES.
package/cluster-glue/Config.in
Quoted 3 lines
package/cluster-glue/Config.in new file +3
-0+ +CLUSTER_GLUE_INSTALL_STAGING = YES +CLUSTER_GLUE_AUTORECONF = YES
Add a comment above saying why it's needed.
package/corosync/Config.in line 2
Quoted 8 lines
package/corosync/Config.in new file +2
-0@@ -0,0 +1,8 @@ +config BR2_PACKAGE_COROSYNC + bool "corosync"
Same question as usual: no dependency on threads? MMU? Any other toolchain feature?
Your .mk file has a dependency on libnss, so you must select it.
package/corosync/Config.in
Quoted 10 lines
package/corosync/Config.in new file +10
-0+################################################################################ +# +# corosync +# +################################################################################ + +COROSYNC_VERSION = 1.4.8 +COROSYNC_SOURCE = corosync-$(COROSYNC_VERSION).tar.gz +COROSYNC_SITE = http://build.clusterlabs.org/corosync/releases +COROSYNC_LICENSE = BSD-3c except for crypto.[ch], public domain for crypto.[ch]
Should be:
BSD-3c (except for crypto.[ch]), Public Domain (crypto.[ch])
package/heartbeat/0001-fix-cross-compilation.patch
Quoted 12 lines
package/corosync/Config.in new file +7
-0+COROSYNC_LICENSE_FILES = LICENSE +COROSYNC_INSTALL_STAGING = YES +COROSYNC_DEPENDENCIES += host-pkgconf libnss + +COROSYNC_CONF_OPTS += ac_cv_func_fnmatch_works=yes + +$(eval $(autotools-package))package/heartbeat/0001-fix-cross-compilation.patch new file +0
-0
Description + SoB and upstream status (ditto for all patches).
package/heartbeat/Config.in line 5
Quoted 11 lines
package/heartbeat/Config.in new file +5
-0@@ -0,0 +1,10 @@ +config BR2_PACKAGE_HEARTBEAT + bool "heartbeat" + select BR2_PACKAGE_UTIL_LINUX + select BR2_PACKAGE_UTIL_LINUX_LIBUUID + select BR2_PACKAGE_IPROUTE2
Replication of dependencies needed here.
package/heartbeat/heartbeat.mk line 5
Quoted 11 lines
package/heartbeat/heartbeat.mk new file +5
-0@@ -0,0 +1,37 @@ +################################################################################ +# +# heartbeat +# +################################################################################
Nit: missing empty line.
package/heartbeat/heartbeat.mk line 11
Quoted 6 lines
package/heartbeat/heartbeat.mk new file +6
-0+HEARTBEAT_VERSION = 958e11be8686 +HEARTBEAT_SOURCE = $(HEARTBEAT_VERSION).tar.bz2 +HEARTBEAT_SITE = http://hg.linux-ha.org/heartbeat-STABLE_3_0/archive +HEARTBEAT_LICENSE = LGPLv2.1+ (library), GPLv2+ (programs) +HEARTBEAT_LICENSE_FILES = doc/COPYING doc/COPYING.LGPL +HEARTBEAT_DEPENDENCIES += host-pkgconf host-python libtool util-linux bzip2 cluster-glue
libtool dependency not needed. Please sync your Config.in with this dependency line by selecting all relevant packages.
Another question: is Python 2.x absolutely needed on the host, or can it work with Python 3.x as well ?
[I'm skipping lots of additional code, for which the same comments appply.]
package/heartbeat/heartbeat.mk line 24
Quoted 13 lines
package/heartbeat/heartbeat.mk new file +13
-0+# This program configure script expects the proper -I flags in CFLAGS and not +# CPPFLAGS +PACEMAKER_CONF_OPTS += \ + CFLAGS="$(TARGET_CFLAGS) -I$(STAGING_DIR)/usr/include -I$(STAGING_DIR)/usr/include/heartbeat" \ + ac_cv_path_HTML2TXT="" \ + ac_cv_path_HELP2MAN="" \ + ac_cv_path_POD2MAN="" \ + ac_cv_path_ASCIIDOC="" \ + ac_cv_path_PUBLICAN="" \ + ac_cv_path_INKSCAPE="" \ + ac_cv_path_XSLTPROC="" \ + ac_cv_path_XMLCATALOG="" \ + ac_cv_path_FOP="" \
We usually just assign to the empty value, and it's normally sufficient.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 23/34 pkg-download: allow git+ssh URLs
In reply to Gilles ChanteperdrixHello,
Quoted 3 lines · Gilles Chanteperdrix
--- package/pkg-download.mk | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
This looks good I believe, so please resend with your SoB line. Maybe an update to the Buildroot manual is also needed.
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 5/34 reproducibility/linux: override build timestamp
In reply to Gilles ChanteperdrixQuoted 10 lines · Gilles Chanteperdrix
with
SOURCE_DATE_EPOCH--- linux/linux.mk | 8 ++++++++ 1 file changed, 8 insertions(+)linux/linux.mk +0
-0
linux/linux.mk LINUX_MAKE_ENV = \ · line 81
Quoted 8 lines
linux/linux.mk +5
-0@@ -74,6 +74,14 @@ LINUX_MAKE_ENV = \ $(TARGET_MAKE_ENV) \ BR_BINARIES_DIR=$(BINARIES_DIR) +ifeq ($(BR2_REPRODUCIBLE),y) +LINUX_MAKE_ENV += \ + KBUILD_BUILD_VERSION=1 \ + KBUILD_BUILD_USER=buildroot \ + KBUILD_BUILD_HOST=buildroot \
This is for example something that really should be inside BR2_REPRODUCIBLE,
because the build user and version gives me useful information while developing.
linux/linux.mk LINUX_MAKE_ENV = \ · line 82
Quoted 1 line
linux/linux.mk +1
-0+ KBUILD_BUILD_TIMESTAMP="$(shell TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH))"
I think the TZ LANG LC_ALL things are redundant, because we already exported
those.
This timestamp is used in several places, so I think it's useful to save it in
a global SOURCE_DATE variable.
Regards,
Arnout
Quoted 6 lines
linux/linux.mk +2
-0+endif + # Get the real Linux version, which tells us where kernel modules are # going to be installed in the target filesystem. LINUX_VERSION_PROBED = `$(MAKE) $(LINUX_MAKE_FLAGS) -C $(LINUX_DIR) --no-print-directory -s kernelrelease 2>/dev/null`
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
Quoted 6 lines · Thomas Petazzoni
Finally, I'm also concerned by the testability of the reproducible build feature. To make sure such a feature work, we would need to do some automated build testing of the same configuration in various environments/configurations and ensure they provide the same output. This is clearly not impossible to do, but there's quite some work involved to set up such an infrastructure.
Even if it is not regularly tested, I think we can start including this
feature. But the help text should clarify that there are no guarantees.
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Gilles ChanteperdrixQuoted 17 lines · Gilles Chanteperdrix
When reproducibility is wanted, generates a global
SOURCE_DATE_EPOCHenvironment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
--- Makefile | 5 +++++ 1 file changed, 5 insertions(+)Makefile +0
-0
Makefile ifeq ($(BR2_REPRODUCIBLE),y) · line 230
Quoted 8 lines
Makefile +5
-0@@ -223,6 +223,11 @@ ifeq ($(BR2_REPRODUCIBLE),y) export TZ=UTC export LANG=C export LC_ALL=C +SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at) +SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s) +export SOURCE_DATE_EPOCH = $(if $(wildcard $(TOPDIR)/.git),$(SOURCE_DATE_GIT),$(SOURCE_DATE_CHANGES))
I really don't like exporting a variable that is not well-known and that
doesn't have a BR_ prefix. Where it is used by packages, it will anyway be in Buildroot-specific patches, so we can just as well add the BR_ prefix.
But actually, we don't want to introduce buildroot-specific (and therefore
unupstreamable) things in patches either. So there is not much point in exporting, except if it's used by scripts.
Regards,
Arnout
Quoted 4 lines
Makefile +0
-0endif # To put more focus on warnings, be less verbose as default
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 7/34 reproducibility/libgcrypt: override timestamps
In reply to Gilles ChanteperdrixQuoted 14 lines · Gilles Chanteperdrix
with
SOURCE_DATE_EPOCH--- package/libgcrypt/0003-honor-source-date-epoch.patch | 15 +++++++++++++++ package/libgcrypt/libgcrypt.mk | 1 + 2 files changed, 16 insertions(+) create mode 100644 package/libgcrypt/0003-honor-source-date-epoch.patchpackage/libgcrypt/0003-honor-source-date-epoch.patch new file +1
-0@@ -0,0 +1,15 @@ +--- libgcrypt-1.6.4/configure.ac.orig 2015-09-08 08:27:26.000000000 +0200
package/libgcrypt/0003-honor-source-date-epoch.patch line 11
Quoted 10 lines
package/libgcrypt/0003-honor-source-date-epoch.patch new file +10
-0++++ libgcrypt-1.6.4/configure.ac 2016-02-23 21:31:19.290079195 +0100 +@@ -1993,7 +1993,11 @@ changequote([,])dnl + BUILD_FILEVERSION="${BUILD_FILEVERSION}mym4_revision_dec" + AC_SUBST(BUILD_FILEVERSION) + +-BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++if test x"${SOURCE_DATE_EPOCH}" = x; then ++ BUILD_TIMESTAMP=`date -u +%Y-%m-%dT%H:%M+0000 2>/dev/null || date` ++else ++ BUILD_TIMESTAMP=`date -u -d @${SOURCE_DATE_EPOCH} +%Y-%m-%dT%H:%M+0000 2>/dev/null || date -d @${SOURCE_DATE_EPOCH}`
Seems to me that SOURCE_DATE_EPOCH is something that is too
package/libgcrypt/libgcrypt.mk LIBGCRYPT_SITE = ftp://ftp.gnupg.org/gcrypt/libgcrypt · line 15
Quoted 13 lines
package/libgcrypt/0003-honor-source-date-epoch.patch new file +4
-0++fi + AC_SUBST(BUILD_TIMESTAMP) + AC_DEFINE_UNQUOTED(BUILD_TIMESTAMP, "$BUILD_TIMESTAMP", + [The time this package was configured for a build])package/libgcrypt/libgcrypt.mk +1
-0@@ -12,6 +12,7 @@ LIBGCRYPT_SITE = ftp://ftp.gnupg.org/gcrypt/libgcrypt LIBGCRYPT_INSTALL_STAGING = YES LIBGCRYPT_DEPENDENCIES = libgpg-error LIBGCRYPT_CONFIG_SCRIPTS = libgcrypt-config +LIBGCRYPT_AUTORECONF = YES
Isn't it sufficient to do
ifeq ($(BR2_REPRODUCIBLE),y) LIBGCRYPT_MAKE_OPTS = BUILD_TIMESTAMP=$(SOURCE_DATE) endif
Regards,
Arnout
Quoted 4 lines
package/libgcrypt/libgcrypt.mk +0
-0LIBGCRYPT_CONF_ENV = \ ac_cv_sys_symbol_underscore=no
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Gilles ChanteperdrixQuoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 10/34 reproducibility/sysstat: override gcc macros
In reply to Gilles ChanteperdrixQuoted 18 lines · Gilles Chanteperdrix
Override the
__DATE__and__TIME__macros with values obtained from the packageSOURCE_DATE_EPOCH.--- package/sysstat/sysstat.mk | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)package/sysstat/sysstat.mk +4
-0@@ -17,6 +17,25 @@ SYSSTAT_DEPENDENCIES += gettext SYSSTAT_MAKE_OPTS += LFLAGS="$(TARGET_LDFLAGS) -lintl" endif +SYSSTAT_CONF_OPTS += \ + sa_lib_dir=/usr/lib/sa + +ifeq ($(BR2_REPRODUCIBLE),y)
package/sysstat/sysstat.mk SYSSTAT_DEPENDENCIES += gettext · line 33
Quoted 10 lines
package/sysstat/sysstat.mk +10
-0+SYSSTAT_SOURCE_DATE = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%b %_d %Y'` +SYSSTAT_SOURCE_TIME = `TZ=UTC LANG=C LC_ALL=C date -d @$(SOURCE_DATE_EPOCH) '+%T'` + +SYSSTAT_CONF_OPTS += \ + CFLAGS="$(TARGET_CFLAGS) -Wno-builtin-macro-redefined -include $(SYSSTAT_DIR)/.br_date_override.h" + +define SYSSTAT_OVERRIDE_DATE_HOOK + echo "#define buildroot_stringify(x) #x" > $(SYSSTAT_DIR)/.br_date_override.h + echo "#define __DATE__ buildroot_stringify($(SYSSTAT_SOURCE_DATE))" >> $(SYSSTAT_DIR)/.br_date_override.h + echo "#define __TIME__ buildroot_stringify($(SYSSTAT_SOURCE_TIME))" >> $(SYSSTAT_DIR)/.br_date_override.h
Actually, we have the toolchain wrapper that could do this override much more
elegantly, and globally for all packages.
Regards,
Arnout
Quoted 9 lines
package/sysstat/sysstat.mk +5
-0+endef + +SYSSTAT_POST_PATCH_HOOKS += SYSSTAT_OVERRIDE_DATE_HOOK +endif + # The isag tool is a post processing script that depends on tcl/tk # among other things. So we don't install it. SYSSTAT_INSTALL_TARGET_OPTS = DESTDIR=$(TARGET_DIR) INSTALL_ISAG=n install
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
Hello,
Quoted 2 lines · Arnout Vandecappelle
Even if it is not regularly tested, I think we can start including this
feature. But the help text should clarify that there are no guarantees.
Yes, absolutely. My question was more about what is the long term plan for this feature. I have no problem with starting to include this feature before we have such a test infrastructure in place, as long as the feature is marked experimental.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Thomas PetazzoniQuoted 5 lines · Thomas Petazzoni
Which has the annoying side effect of increasing even more the set of mandatory dependencies to build libglib :-/
So I wonder if the dependency on host-python should be made conditional on
BR2_REPRODUCIBLE.
Hi.
Generally, there's no use case of having gdbus-codegen in the target since it's used at build time.
Quoted 5 lines
Since then, Gustavo has already modified the libglib2 package to remove the
/usr/share/glib-2.0/codegendirectory. However, gdbus-codegen is not removed.See commit
1f1f16e9e5367048faa1cf17237e0c0d422e98d6.
Exactly, gdbus-codegen is a leftover that i missed, it can be safely removed.
Regards.
PATCH 9/34 reproducibility/libglib2: allow removing codegen
Hello,
Quoted 2 lines · Gustavo Zacarias
Exactly, gdbus-codegen is a leftover that i missed, it can be safely removed.
Thanks for confirming. Could you submit a patch doing this?
Thanks!
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 13/34 reproducibility/libtool: fix reproducibility issues
In reply to Thomas PetazzoniQuoted 28 lines · Thomas Petazzoni
Hello,
This probably should be several separate patches.
package/libtool/0001-fix-cross-compilation.patch
Quoted 16 lines · Gilles Chanteperdrix
- remove libtool and libtoolize; they are full of references to the build environment tools, and unlikely to be needed on target;
- do not use/etc/ld.so.confcontents in libtool search path: they are not relevant on target, and cause a useless dependency of libltdl on the build environment.--- package/libtool/0001-fix-cross-compilation.patch | 11 +++++++++++ package/libtool/libtool.mk | 8 ++++++-- 2 files changed, 17 insertions(+), 2 deletions(-) create mode 100644 package/libtool/0001-fix-cross-compilation.patchpackage/libtool/0001-fix-cross-compilation.patch new file +0
-0This patch should have a description + Signed-off-by.
However, libtool is really a core package, and not a simple one, so I'm a bit uneasy with the idea of carrying patches against it. What are the chances of having something like this submitted upstream?
Quoted 13 lines
package/libtool/libtool.mk LIBTOOL_VERSION = 2.4.6 · line 11
Quoted 11 lines
package/libtool/0001-fix-cross-compilation.patch new file +1
-0+ fipackage/libtool/libtool.mk +1
-1@@ -8,7 +8,7 @@ LIBTOOL_VERSION = 2.4.6 LIBTOOL_SOURCE = libtool-$(LIBTOOL_VERSION).tar.xz LIBTOOL_SITE = $(BR2_GNU_MIRROR)/libtool LIBTOOL_INSTALL_STAGING = YES -LIBTOOL_CONF_ENV = HELP2MAN=true +LIBTOOL_CONF_OPTS = HELP2MAN=trueWhy?
Well, passing autoconf configure arguments on its command line and not in its environment has been the recommended way for something like 16 years. It is recommended in the "goat book", published in 2000.
If you pass the argument on the command line, its gets built into config.status, so that if config.status needs to be run again, the option gets passed to configure again. On the other hand, if you use the environment, if config.status is not run in an environment where the option is defined, the option gets forgotten.
We patch the autotools files and (have to) run autoreconf, and for a reason of file dates I forgot, when we run make, config.status is re-run without the LIBTOOL_CONV_ENV, and the HELP2MAN variable gets forgotten.
Quoted 11 lines
package/libtool/libtool.mk define LIBTOOL_AVOID_AUTORECONF_HOOK · line 36
Quoted 9 lines
package/libtool/libtool.mk +0
-1LIBTOOL_DEPENDENCIES = host-m4 LIBTOOL_LICENSE = GPLv2+ LIBTOOL_LICENSE_FILES = COPYING@@ -33,9 +33,13 @@ define LIBTOOL_AVOID_AUTORECONF_HOOK find $(@D) -name configure -exec touch '{}' \; find $(@D) -name Makefile.in -exec touch '{}' \; endef -LIBTOOL_PRE_CONFIGURE_HOOKS += LIBTOOL_AVOID_AUTORECONF_HOOKWhy?
Because the patch touches the autoconf files.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 4/34 reproducibility: make rootfs.tar reproducible
In reply to Arnout VandecappelleQuoted 24 lines · Arnout Vandecappelle
Quoted 13 lines · Thomas Petazzoni
Hello,
fs/tar/tar.mk line 19
Quoted 6 lines · Gilles Chanteperdrix
fs/tar/tar.mk +6
-0+define ROOTFS_TAR_CMD + cd $(TARGET_DIR) && { \ + find . -\( -! -type d -o -empty -\) -print0 | \ + sort -z | \ + tar --null -T - -c$(TAR_OPTS)f $@ --mtime=@$(SOURCE_DATE_EPOCH) --numeric-owner; \ + }We normally write such constructs as:
Quoted 2 lines
(cd $(TARGET_DIR) && \ foo ....)Actually, we don't AFAIK... In general, the parenthesis are not needed so they
should be removed. So also in this case it should becd $(TARGET_DIR); find ....Quoted 1 line
Quoted 7 lines
Quoted 4 lines
However, this raises the question of what's needed for all the other filesystem formats. Will they all have to implement a different ROOTFS_<foo>_CMD variable ? Or will there be some commonalities that should be factored out in the common rootfs image infrastructure ?
Yes, to me it makes more sense to do this in the actual target directory after
the post-build scripts have been run.
I do not understand what you mean. The aim of the command is to sort the list of files passed to tar, it does not operate on the file system.
Sorting is indeed needed for most other outputs I have tested (cpio, isofs).
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 19/34 reproducibility/python: make python reproducible
In reply to Thomas PetazzoniQuoted 11 lines · Thomas Petazzoni
Hello,
Quoted 7 lines · Gilles Chanteperdrix
python build is not reproducible for two reasons:
- the libpython library embeds the build timestamps, fix this by overriding the__DATE__and__TIME__macro usingSOURCE_DATE_EPOCH;
-_sysconfigdata.pyand config/Makefile on target includes references through INSTALL andMKDIR_Pvariables to host tools, fix this by overriding these variables values with the corresponding paths on buildroot rootfs.
Quoted 3 lines
For this second part, are you really aiming at having reproducible builds with different build directories ? As you say in your cover letter, that's an even more complicated problem.
No, the aim is to get binaries which do not depend on the location of "install" and "mkdir" on the host file system, but rather get them to depend on their location on the target file system, since it is where theses files are going to be used.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 22/34 High availability: add packages
In reply to Thomas PetazzoniQuoted 10 lines · Thomas Petazzoni
Hello,
Quoted 5 lines · Gilles Chanteperdrix
cluster-glue
corosync
heartbeat
libqb
pacemakerThis should be split in a series of patches, with one package per patch.
Please ignore this patch, the packages are not even fully integrated and will not start automatically on boot. I did not intend to post it.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Arnout VandecappelleQuoted 22 lines · Arnout Vandecappelle
Quoted 21 lines · Gilles Chanteperdrix
When reproducibility is wanted, generates a global
SOURCE_DATE_EPOCHenvironment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
--- Makefile | 5 +++++ 1 file changed, 5 insertions(+)Makefile +1
-0@@ -223,6 +223,11 @@ ifeq ($(BR2_REPRODUCIBLE),y) export TZ=UTC export LANG=C export LC_ALL=C +SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at)
Quoted 8 lines
Makefile ifeq ($(BR2_REPRODUCIBLE),y) · line 230
Quoted 4 lines
Makefile +4
-0+SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s) +export SOURCE_DATE_EPOCH = $(if $(wildcard $(TOPDIR)/.git),$(SOURCE_DATE_GIT),$(SOURCE_DATE_CHANGES))I really don't like exporting a variable that is not well-known and that
doesn't have aBR_prefix. Where it is used by packages, it will anyway be in Buildroot-specific patches, so we can just as well add theBR_prefix.
SOURCE_DATE_EPOCH is well-known. Its meaning is defined by a specification:
https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal
Quoted 4 lines
But actually, we don't want to introduce buildroot-specific (and therefore
unupstreamable) things in patches either. So there is not much point in exporting, except if it's used by scripts.
I used it in some patched configure scripts (libnspr, if I remember correctly). Which is why I exported it.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Arnout VandecappelleQuoted 6 lines · Arnout Vandecappelle
Quoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?
I am not a python specialist, but it seems to me four bytes in the python generated bytecode are the build timestamp, so unless there is a way to override it with SOURCE_DATE_EPOCH, I do not see that possible.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 11/34 reproducibility/vim: make vim reproducible
In reply to Thomas PetazzoniQuoted 7 lines · Thomas Petazzoni
package/vim/vim.mk else · line 57
Quoted 5 lines
package/vim/vim.mk +5
-0+define VIM_FIX_AWK_REFERENCE + sed -i 's,^#!.*awk,#!/usr/bin/awk,' $(TARGET_DIR)/usr/share/vim/vim74/tools/mve.awk +endef + +VIM_POST_INSTALL_TARGET_HOOKS += VIM_FIX_AWK_REFERENCEDo we even need this awk script installed?
How should I know, I am an emacs user...
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 17/34 reproducibility/libnspr: override build timestamps
In reply to Thomas PetazzoniQuoted 12 lines · Thomas Petazzoni
Hello,
Quoted 1 line · Gilles Chanteperdrix
with
SOURCE_DATE_EPOCHMissing SoB line here.
Quoted 5 lines
--- package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch | 71 ++++++++++++++++++++++ 1 file changed, 71 insertions(+) create mode 100644 package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch
Quoted 8 lines
package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch
Quoted 5 lines
package/libnspr/0003-honor-SOURCE_DATE_EPOCH.patch new file +0
-0Missing patch description and Signed-off-by line. Also, did you submit this upstream?
No, it looks a bit specific to me.
Quoted 1 line
Does Debian carry a similar patch for its Reproducible Builds effort?
No idea.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 23/34 pkg-download: allow git+ssh URLs
In reply to Thomas PetazzoniQuoted 10 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Thomas" == Thomas Petazzoni <[email protected]> writes:
Hello,
Quoted 3 lines · Gilles Chanteperdrix
---
package/pkg-download.mk| 2 +-1 file changed, 1 insertion(+), 1 deletion(-)This looks good I believe, so please resend with your SoB line. Maybe an update to the Buildroot manual is also needed.
I'm not sure we want to advertise this feature too much (or at least mention it is deprecated). I've personally not heard about git+ssh://, and googling around I didn't find anything right away.
The only reference I have found is in the git source code:
static enum protocol get_protocol(const char *name)
{
if (!strcmp(name, "ssh"))
return PROTO_SSH;
if (!strcmp(name, "git"))
return PROTO_GIT;
if (!strcmp(name, "git+ssh")) /* deprecated - do not use */
return PROTO_SSH;
if (!strcmp(name, "ssh+git")) /* deprecated - do not use */
return PROTO_SSH;
if (!strcmp(name, "file"))
return PROTO_FILE;
die("I don't handle protocol '%s'", name);
}So the question is how long git will continue to support it.
With that said, I don't have a problem with us detecting git+ssh:// (and perhaps ssh+git://) as git URLs and forwarding it to the download wrapper - But if we start documenting it as supported we should continue to do so, so we might end up having to rewrite git+ssh:// -> ssh:// URLs and so on when/if git stops supporting them natively.
Signature · 1 line
Bye, Peter Korsgaard
PATCH 21/34 hddtemp: add
In reply to Gilles ChanteperdrixQuoted 17 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Gilles" == Gilles Chanteperdrix <[email protected]> writes:
--- package/Config.in | 1 + package/hddtemp/Config.in | 8 ++++++++ package/hddtemp/hddtemp.hash | 2 ++ package/hddtemp/hddtemp.mk | 19 +++++++++++++++++++ 4 files changed, 30 insertions(+) create mode 100644 package/hddtemp/Config.in create mode 100644 package/hddtemp/hddtemp.hash create mode 100644 package/hddtemp/hddtemp.mkpackage/Config.in +0
-0
package/Config.in comment "Utilities" · line 1461
Quoted 4 lines
package/Config.in +1
-0@@ -1458,6 +1458,7 @@ comment "Utilities" source "package/file/Config.in" source "package/gnupg/Config.in" source "package/gnupg2/Config.in" + source "package/hddtemp/Config.in"
I think this should rather go under "Hardware handling" like smartmontools or hdparm.
Quoted 9 lines
package/Config.in +0
-0source "package/inotify-tools/Config.in" source "package/lockfile-progs/Config.in" source "package/logrotate/Config.in"package/hddtemp/Config.in new file +0
-0
package/hddtemp/Config.in line 6
Quoted 6 lines
package/hddtemp/Config.in new file +6
-0@@ -0,0 +1,8 @@ +config BR2_PACKAGE_HDDTEMP + bool "hddtemp" + help + hddtemp is a small utility that gives you the temperature of your + hard drive by reading S.M.A.R.T. informations, for drives that + support this feature.
It also uses iconv (to figure out how to print a degrees celsius sign, gaah), so you need to add:
select BR2_PACKAGE_LIBICONV if !BR2_ENABLE_LOCALE
And add it to HDDTEMP_DEPENDENCIES. A quick test shows that it forgets to link with it, so it needs some help:
ifeq ($(BR2_PACKAGE_LIBICONV),y)HDDTEMP_DEPENDENCIES += libiconvHDDTEMP_CONF_ENV += LIBS="-liconv"
endif
It also needs a small patch to build without gettext as it unconditionally calls:
setlocale (LC_ALL, ""); bindtextdomain (PACKAGE, LOCALEDIR); textdomain (PACKAGE);
Those needs to be inside #ifdef ENABLE_NLS conditionals.
And finally there's code in backtrace.c that breaks when building with uClibc on i386 (as uClibc also defines __GLIBC__ for compatibility but doesn't provide backtrace functionality by default):
#if defined(__i386__) && defined(__GLIBC__)
These patches should preferably be upstreamed, but as the latest (beta!) release is 10 years old by now, the project seems pretty dead.
The savannah status is also listed as Orphaned/Unmaintained.
Quoted 16 lines
package/hddtemp/Config.in new file +2
-0+ + http://www.guzu.net/linux/hddtemp.phppackage/hddtemp/hddtemp.hash new file +2
-0@@ -0,0 +1,2 @@ +# Locally calculated after checking pgp signature +sha256 618541584054093d53be8a2d9e81c97174f30f00af91cb8700a97e442d79ef5b hddtemp-0.3-beta15.tar.bz2package/hddtemp/hddtemp.mk new file +0
-0
package/hddtemp/hddtemp.mk line 10
Quoted 10 lines
package/hddtemp/hddtemp.mk new file +10
-0@@ -0,0 +1,19 @@ +################################################################################ +# +# hddtemp +# +################################################################################ + +HDDTEMP_VERSION = 0.3-beta15 +HDDTEMP_SOURCE = hddtemp-$(HDDTEMP_VERSION).tar.bz2 +HDDTEMP_SITE = http://download.savannah.gnu.org/releases/hddtemp +HDDTEMP_LICENSE = GPLv2
The source files contain the "(at your option) any later version" wording, so this should be GPLv2+
package/hddtemp/hddtemp.mk line 15
Quoted 5 lines
package/hddtemp/hddtemp.mk new file +5
-0+HDDTEMP_LICENSE_FILES = GPL-2 + +define HDDTEMP_REMOVE_MAN + rm -rf $(TARGET_DIR)/usr/man +endef
Like Thomas mentioned, this is already taken care of in target-finalize.
The big question is how to handle hddtemp.db?
The tarball doesn't contain the file, so hddtemp doesn't do anything sensible:
sudo ./target/usr/sbin/hddtemp /dev/sda
hddtemp: can't open /usr/share/misc/hddtemp.db: No such file or directory
The file is available on the upstream website, but it is not versioned - So that is not suitable for Buildroot:
http://www.guzu.net/linux/hddtemp.db
Either we:
- Convince upstream to version this file (and preferably include it in the tarball) - But as the project seems dead this might be hard
- Apply the Debian patch which includes (an older version of?) hddtemp.db. Notice that Debian also patches various things in the code, so please check that the changes make sense in the context of Buildroot:
http://http.debian.net/debian/pool/main/h/hddtemp/hddtemp_0.3-beta15-52.diff.gz
- We include hddtemp.db in package/hddtemp. It isn't really nice, but the file is relatively small (29K)
Given the missing hddtemp.db, how have you tested this package?
Signature · 1 line
Bye, Peter Korsgaard
PATCH 1/34 reproducibility: introduce config knob
In reply to Gilles ChanteperdrixQuoted 14 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Gilles" == Gilles Chanteperdrix <[email protected]> writes:
Making the builds reproducible involves turning off build timestamps, which some users may find annoying. So make the reproducible builds optional.
--- Config.in | 3 +++ 1 file changed, 3 insertions(+)Config.in +0
-0
Config.in config BR2_COMPILER_PARANOID_UNSAFE_PATH · line 628
Quoted 6 lines
Config.in +3
-0@@ -623,6 +623,9 @@ config BR2_COMPILER_PARANOID_UNSAFE_PATH endmenu +config BR2_REPRODUCIBLE + bool "Try and remove all sources of non-reproducibility from the build" +
Hmm, I'm not sure I like the extra complexity of 2 ways of doing things. What are the use cases for these timestamps? Are they important enough to keep?
In any case, as enabling this option makes us do "more" (E.G. pass extra arguments to tools), we should imho enable it by default to ensure those code paths gets tested.
Signature · 1 line
Bye, Peter Korsgaard
PATCH 1/34 reproducibility: introduce config knob
Hello,
Quoted 3 lines · Peter Korsgaard
Hmm, I'm not sure I like the extra complexity of 2 ways of doing things. What are the use cases for these timestamps? Are they important enough to keep?
I use timestamps all the time to verify that I'm indeed running the version of the code I intend to run. As I work on the kernel, I use the kernel timestamp, but I really use it all the time, to double check that I'm running the kernel image I just built and not one that was left around, built several days/weeks ago.
So timestamps are useful IMO.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 1/34 reproducibility: introduce config knob
Quoted 9 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Thomas" == Thomas Petazzoni <[email protected]> writes:
Hello,
Quoted 3 lines · Peter Korsgaard
Hmm, I'm not sure I like the extra complexity of 2 ways of doing things. What are the use cases for these timestamps? Are they important enough to keep?
Quoted 5 lines
I use timestamps all the time to verify that I'm indeed running the version of the code I intend to run. As I work on the kernel, I use the kernel timestamp, but I really use it all the time, to double check that I'm running the kernel image I just built and not one that was left around, built several days/weeks ago.
Ok, what timestamps exactly? Embedded timestamps like uname -v or filesystem timestamps of the files?
Signature · 2 lines
Venlig hilsen, Peter Korsgaard
PATCH 1/34 reproducibility: introduce config knob
Hello,
Quoted 8 lines · Peter Korsgaard
Quoted 5 lines
I use timestamps all the time to verify that I'm indeed running the version of the code I intend to run. As I work on the kernel, I use the kernel timestamp, but I really use it all the time, to double check that I'm running the kernel image I just built and not one that was left around, built several days/weeks ago.
Ok, what timestamps exactly? Embedded timestamps like uname -v or filesystem timestamps of the files?
The one I use for the kernel is obviously the embedded time stamp, displayed at boot time (and also in uname -v).
I guess the filesystem time stamps are not that useful, but they are the "easy" part (i.e they can be fixed globally when generating the filesystem image). The embedded timestamps all over the place in different programs/libraries are really the annoying part.
Thomas
Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com
PATCH 23/34 pkg-download: allow git+ssh URLs
In reply to Peter KorsgaardQuoted 8 lines · Peter Korsgaard
Quoted 7 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Thomas" == Thomas Petazzoni <[email protected]> writes:
Hello,
Quoted 3 lines · Gilles Chanteperdrix
---
package/pkg-download.mk| 2 +-1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 6 lines
Quoted 2 lines
This looks good I believe, so please resend with your SoB line. Maybe an update to the Buildroot manual is also needed.
I'm not sure we want to advertise this feature too much (or at least mention it is deprecated). I've personally not heard about git+ssh://, and googling around I didn't find anything right away.
Well, I always used git+ssh://, I did not even know it was deprecated. The advantage of git+ssh is that the scheme clearly identifies git over ssh, so we can choose the git helper. If we got the ssh scheme to do the same and other revision control systems used ssh URL (I do not even know, I have not been using something else than git for a long time), it is not going to be possible to find out which helper to use.
One way or another, I believe it is helpful to be able to use git over ssh URLs, are they may be used for projects specific to a company, not public.
Regards.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 21/34 hddtemp: add
In reply to Peter KorsgaardQuoted 64 lines · Peter Korsgaard
Quoted 17 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
"Gilles" == Gilles Chanteperdrix <[email protected]> writes:
--- package/Config.in | 1 + package/hddtemp/Config.in | 8 ++++++++ package/hddtemp/hddtemp.hash | 2 ++ package/hddtemp/hddtemp.mk | 19 +++++++++++++++++++ 4 files changed, 30 insertions(+) create mode 100644 package/hddtemp/Config.in create mode 100644 package/hddtemp/hddtemp.hash create mode 100644 package/hddtemp/hddtemp.mkpackage/Config.in +0
-0package/Config.in comment "Utilities" · line 1461
Quoted 4 lines
package/Config.in +1
-0@@ -1458,6 +1458,7 @@ comment "Utilities" source "package/file/Config.in" source "package/gnupg/Config.in" source "package/gnupg2/Config.in" + source "package/hddtemp/Config.in"I think this should rather go under "Hardware handling" like smartmontools or hdparm.
Quoted 9 lines
package/Config.in +0
-0source "package/inotify-tools/Config.in" source "package/lockfile-progs/Config.in" source "package/logrotate/Config.in"package/hddtemp/Config.in new file +0
-0package/hddtemp/Config.in line 6
Quoted 6 lines
package/hddtemp/Config.in new file +6
-0@@ -0,0 +1,8 @@ +config BR2_PACKAGE_HDDTEMP + bool "hddtemp" + help + hddtemp is a small utility that gives you the temperature of your + hard drive by reading S.M.A.R.T. informations, for drives that + support this feature.It also uses iconv (to figure out how to print a degrees celsius sign, gaah), so you need to add:
select
BR2_PACKAGE_LIBICONVif !BR2_ENABLE_LOCALEAnd add it to
HDDTEMP_DEPENDENCIES. A quick test shows that it forgets to link with it, so it needs some help:ifeq ($(BR2_PACKAGE_LIBICONV),y)
HDDTEMP_DEPENDENCIES+= libiconvHDDTEMP_CONF_ENV+= LIBS="-liconv"
endifIt also needs a small patch to build without gettext as it unconditionally calls:
setlocale (LC_ALL, ""); bindtextdomain (PACKAGE, LOCALEDIR); textdomain (PACKAGE);Those needs to be inside #ifdef
ENABLE_NLSconditionals.
Quoted 8 lines
And finally there's code in
backtrace.cthat breaks when building with uClibc on i386 (as uClibc also defines__GLIBC__for compatibility but doesn't provide backtrace functionality by default):#if
defined(__i386__)&&defined(__GLIBC__)These patches should preferably be upstreamed, but as the latest (beta!) release is 10 years old by now, the project seems pretty dead.
Ok, let us forget about this patch, as for the "high-availability" patch, I did not expect to post this patch with the rest, it landed here only because I posted the patch list, unsorted.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Gilles ChanteperdrixQuoted 32 lines · Gilles Chanteperdrix
Quoted 22 lines · Arnout Vandecappelle
Quoted 21 lines · Gilles Chanteperdrix
When reproducibility is wanted, generates a global
SOURCE_DATE_EPOCHenvironment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
--- Makefile | 5 +++++ 1 file changed, 5 insertions(+)Makefile +1
-0@@ -223,6 +223,11 @@ ifeq ($(BR2_REPRODUCIBLE),y) export TZ=UTC export LANG=C export LC_ALL=C +SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at)Quoted 8 lines
Makefile ifeq ($(BR2_REPRODUCIBLE),y) · line 230
Quoted 4 lines
Makefile +4
-0+SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s) +export SOURCE_DATE_EPOCH = $(if $(wildcard $(TOPDIR)/.git),$(SOURCE_DATE_GIT),$(SOURCE_DATE_CHANGES))I really don't like exporting a variable that is not well-known and that
doesn't have aBR_prefix. Where it is used by packages, it will anyway be in Buildroot-specific patches, so we can just as well add theBR_prefix.
Quoted 3 lines
SOURCE_DATE_EPOCHis well-known. Its meaning is defined by a specification:
https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal
Ah, good.
Actually, a better reference is
https://reproducible-builds.org/specs/source-date-epoch/
(actually the same developers wrote this spec, but it's supposedly not debian-specific).
When you repost, could you make sure that the commit message mentions this
explicitly?
Quoted 8 lines
Quoted 5 lines
But actually, we don't want to introduce buildroot-specific (and therefore
unupstreamable) things in patches either. So there is not much point in exporting, except if it's used by scripts.I used it in some patched configure scripts (libnspr, if I remember correctly). Which is why I exported it.
As I said: used in patches.
But indeed, it is well-known. So OK to export it.
BTW, when it is possible to download a debian patch for a package (by settingFOO_PATCH = ...) rather than rolling our own, we prefer that.
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 4/34 reproducibility: make rootfs.tar reproducible
In reply to Gilles ChanteperdrixQuoted 30 lines · Gilles Chanteperdrix
Quoted 29 lines · Arnout Vandecappelle
Quoted 13 lines · Thomas Petazzoni
Hello,
fs/tar/tar.mk line 19
Quoted 6 lines · Gilles Chanteperdrix
fs/tar/tar.mk +6
-0+define ROOTFS_TAR_CMD + cd $(TARGET_DIR) && { \ + find . -\( -! -type d -o -empty -\) -print0 | \ + sort -z | \ + tar --null -T - -c$(TAR_OPTS)f $@ --mtime=@$(SOURCE_DATE_EPOCH) --numeric-owner; \ + }We normally write such constructs as:
Quoted 2 lines
(cd $(TARGET_DIR) && \ foo ....)Actually, we don't AFAIK... In general, the parenthesis are not needed so they
should be removed. So also in this case it should becd $(TARGET_DIR); find ....Quoted 5 lines
However, this raises the question of what's needed for all the other filesystem formats. Will they all have to implement a different ROOTFS_<foo>_CMD variable ? Or will there be some commonalities that should be factored out in the common rootfs image infrastructure ?
Quoted 6 lines
Quoted 2 lines
Yes, to me it makes more sense to do this in the actual target directory after
the post-build scripts have been run.I do not understand what you mean. The aim of the command is to sort the list of files passed to tar, it does not operate on the file system.
Duh, sorry, I was just talking about the --mtime bit.
Regards,
Arnout
Quoted 4 lines
Sorting is indeed needed for most other outputs I have tested (cpio, isofs).
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Gilles ChanteperdrixQuoted 5 lines · Gilles Chanteperdrix
Quoted 4 lines · Arnout Vandecappelle
Quoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Quoted 7 lines
Quoted 2 lines
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?I am not a python specialist, but it seems to me four bytes in the python generated bytecode are the build timestamp, so unless there is a way to override it with
SOURCE_DATE_EPOCH, I do not see that possible.
I've checked the docs. What is saved is the timestamp of the .py file, not the
build time.
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.
Regards,
Arnout
[1] https://lists.fedoraproject.org/pipermail/python-devel/2014-August/000617.html
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 9/34 reproducibility/libglib2: allow removing codegen
Quoted 9 lines · Arnout Vandecappelle
Quoted 8 lines · Gilles Chanteperdrix
Quoted 6 lines · Arnout Vandecappelle
Quoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?
Quoted 7 lines
Quoted 4 lines
I am not a python specialist, but it seems to me four bytes in the python generated bytecode are the build timestamp, so unless there is a way to override it with
SOURCE_DATE_EPOCH, I do not see that possible.I've checked the docs. What is saved is the timestamp of the .py file, not the
build time.
Mmmm I don't remember. I would run a compilation manually, twice at a different time, to make sure that the problem is only the file date.
Quoted 5 lines
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.
I think I tried running "touch" before compiling, unfortunately playing with file dates before running "make" is a bit like playing with fire. For instance with autotools based projects for which autoreconf is not run, the project may use versions of the autotools not installed on the user machine, and because of file dates may want to rerun autoconf or autmake and whine because the right versions are not available. Doing this only for .py files is much more reasonable.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 3/34 reproducibility: generate SOURCE_DATE_EPOCH
In reply to Arnout VandecappelleQuoted 56 lines · Arnout Vandecappelle
Quoted 32 lines · Gilles Chanteperdrix
Quoted 22 lines · Arnout Vandecappelle
Quoted 21 lines · Gilles Chanteperdrix
When reproducibility is wanted, generates a global
SOURCE_DATE_EPOCHenvironment variable which contains either the date of buildroot last commit if running from a git repository, or the latest release date.This means that all packages embedding build dates will appear to have the same build date, so in case of new commit or release, all packages will appear to have been change, even though some of them may not have changed in fact.
--- Makefile | 5 +++++ 1 file changed, 5 insertions(+)Makefile +1
-0@@ -223,6 +223,11 @@ ifeq ($(BR2_REPRODUCIBLE),y) export TZ=UTC export LANG=C export LC_ALL=C +SOURCE_DATE_GIT = $(shell GIT_DIR=$(TOPDIR)/.git $(GIT) log -1 --format=%at)Quoted 8 lines
Makefile ifeq ($(BR2_REPRODUCIBLE),y) · line 230
Quoted 4 lines
Makefile +4
-0+SOURCE_DATE_CHANGES = $(shell head -n 1 $(TOPDIR)/CHANGES | \ + sed 's/^.*Released \(.*\)$$/\1/;s/\(st\|nd\|rd\|th\),//' | \ + LANG=C LC_ALL=C TZ=UTC xargs -i date -d \{\} +%s) +export SOURCE_DATE_EPOCH = $(if $(wildcard $(TOPDIR)/.git),$(SOURCE_DATE_GIT),$(SOURCE_DATE_CHANGES))I really don't like exporting a variable that is not well-known and that
doesn't have aBR_prefix. Where it is used by packages, it will anyway be in Buildroot-specific patches, so we can just as well add theBR_prefix.Quoted 3 lines
SOURCE_DATE_EPOCHis well-known. Its meaning is defined by a specification:
https://wiki.debian.org/ReproducibleBuilds/TimestampsProposalAh, good.
Actually, a better reference is
https://reproducible-builds.org/specs/source-date-epoch/
(actually the same developers wrote this spec, but it's supposedly not debian-specific).When you repost, could you make sure that the commit message mentions this
explicitly?Quoted 8 lines
Quoted 5 lines
But actually, we don't want to introduce buildroot-specific (and therefore
unupstreamable) things in patches either. So there is not much point in exporting, except if it's used by scripts.I used it in some patched configure scripts (libnspr, if I remember correctly). Which is why I exported it.
Quoted 6 lines
As I said: used in patches.
But indeed, it is well-known. So OK to export it.
BTW, when it is possible to download a debian patch for a package (by setting
FOO_PATCH= ...) rather than rolling our own, we prefer that.
The thing is that trying and understanding what Debian does is not convenient for me and most of the time a waste of time:
- things that I believe are problems for buildroot may not be considered problems for Debian (see for instance the libltdl issue, or the grub2 issue).
- the Debian wiki has a high noise/signal ratio, and is not always up to date, and this when I can access it, because the IP addresses I use for browsing are blacklisted by Debian, so I have to take special steps to access it.
So, I gave up on Debian pretty early, and I am not interested in trying that painful exercise again. If someone else wants to do it though, go ahead.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Gilles ChanteperdrixQuoted 15 lines · Gilles Chanteperdrix
Quoted 14 lines · Arnout Vandecappelle
Quoted 12 lines · Gilles Chanteperdrix
Quoted 6 lines · Arnout Vandecappelle
Quoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?I am not a python specialist, but it seems to me four bytes in the python generated bytecode are the build timestamp, so unless there is a way to override it with
SOURCE_DATE_EPOCH, I do not see that possible.
Quoted 6 lines
Quoted 2 lines
I've checked the docs. What is saved is the timestamp of the .py file, not the
build time.Mmmm I don't remember. I would run a compilation manually, twice at a different time, to make sure that the problem is only the file date.
I did that - admittedly with just a few seconds difference. Both in
python2.7.11 and python3.5, they were identical when compiling a second time, and different after touch'ing.
Quoted 7 lines
Quoted 6 lines
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.
Quoted 8 lines
I think I tried running "touch" before compiling, unfortunately playing with file dates before running "make" is a bit like playing with fire. For instance with autotools based projects for which autoreconf is not run, the project may use versions of the autotools not installed on the user machine, and because of file dates may want to rerun autoconf or autmake and whine because the right versions are not available. Doing this only for .py files is much more reasonable.
I tested this as well, with make-3.81 and make-4.04. Both of them do _not_
rebuild if the timestamps are identical. And since the idea is to use touch -d @$(SOURCE_DATE_EPOCH), all timestamps will be identical.
But it does indeed mean that if a package has a generated file with an earlier
date than the source files, it will now suddenly no longer be rebuilt.
I don't know how the various other build systems (waf, scons) will behave.
There is indeed a risk, but I'd say we deal with that when it happens. Actually I think we would prefer to detect such issues.
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 16/34 reproducibility/lzo: forbid lzo with reproducible builds
In reply to Thomas PetazzoniQuoted 35 lines · Thomas Petazzoni
Hello,
Quoted 1 line · Gilles Chanteperdrix
The lzo archive contains a timestamp.
Missing Signed-off-by line.
Quoted 17 lines
--- fs/tar/Config.in | 7 +++++++ 1 file changed, 7 insertions(+)fs/tar/Config.in +1
-0@@ -33,6 +33,7 @@ config BR2_TARGET_ROOTFS_TAR_LZMA config BR2_TARGET_ROOTFS_TAR_LZO bool "lzo" + depends on !BR2_REPRODUCIBLE help Do compress the tarball with lzop.fs/tar/Config.in config BR2_TARGET_ROOTFS_TAR_XZ · line 51
Quoted 8 lines
fs/tar/Config.in +5
-0@@ -43,6 +44,12 @@ config BR2_TARGET_ROOTFS_TAR_XZ endchoice +if BR2_REPRODUCIBLE + +comment "lzo compression is only available with non-reproducible builds" + +endifWhen we are in such situations, we normally do:
Quoted 7 lines
comment "blablabla"
depends onBR2_REPRODUCIBLEhowever, in this specific case, I would simply not bother adding a comment. If we start adding a comment for everything that is not suitable for reproducible builds, we're going to add hundreds of them I believe.
IMHO, it should be different: things that are not reproducible should not be
excluded, but instead a comment should be shown to warn the user the build is not reproducible. Perhaps like for legacy: a config option that is not reproducible selects BR2_NOT_REPRODUCIBLE (a blind option), and at the bottom of the top level menu we add a comment that depends on BR2_REPRODUCIBLE && BR2_NOT_REPRODUCIBLE. It will be a bit tricky to find out exactly which option is the one that triggers this comment, but it's doable by searching for BR2_NOT_REPRODUCIBLE and seeing which one in the long list of selectors is y.
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Arnout VandecappelleQuoted 43 lines · Arnout Vandecappelle
Quoted 15 lines · Gilles Chanteperdrix
Quoted 14 lines · Arnout Vandecappelle
Quoted 12 lines · Gilles Chanteperdrix
Quoted 6 lines · Arnout Vandecappelle
Quoted 2 lines · Gilles Chanteperdrix
But this is not sufficient, compiling the python bytecode with the same interpreter in different environments yields different binaries,
Er, this is worrisome... You are saying that we don't have a chance in hell of
generating reproducible python bytecode?I am not a python specialist, but it seems to me four bytes in the python generated bytecode are the build timestamp, so unless there is a way to override it with
SOURCE_DATE_EPOCH, I do not see that possible.Quoted 6 lines
Quoted 2 lines
I've checked the docs. What is saved is the timestamp of the .py file, not the
build time.Mmmm I don't remember. I would run a compilation manually, twice at a different time, to make sure that the problem is only the file date.
I did that - admittedly with just a few seconds difference. Both in
python2.7.11 and python3.5, they were identical when compiling a second time, and different after touch'ing.Quoted 7 lines
Quoted 6 lines
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.Quoted 8 lines
I think I tried running "touch" before compiling, unfortunately playing with file dates before running "make" is a bit like playing with fire. For instance with autotools based projects for which autoreconf is not run, the project may use versions of the autotools not installed on the user machine, and because of file dates may want to rerun autoconf or autmake and whine because the right versions are not available. Doing this only for .py files is much more reasonable.
Quoted 3 lines
I tested this as well, with make-3.81 and make-4.04. Both of them do _not_
rebuild if the timestamps are identical. And since the idea is to use touch -d @$(SOURCE_DATE_EPOCH), all timestamps will be identical.
You checked both the pyo and the pyc?
Quoted 4 lines
But it does indeed mean that if a package has a generated file with an earlier
date than the source files, it will now suddenly no longer be rebuilt.
Yes, that is another problem. But I tried it, and this is not the one I had, the one I had was the contrary: the dates made make want to rebuild some files (the autotools/automake files), whereas the right versions of autoconf and automake were not installed, this was a package that did not run autoreconf. The "make" tool is completely based on file dates, so again, I think messing with file dates before running make is a bad idea.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 9/34 reproducibility/libglib2: allow removing codegen
Quoted 3 lines · Gilles Chanteperdrix
Quoted 2 lines · Arnout Vandecappelle
Quoted 1 line · Gilles Chanteperdrix
On Tue, May 10, 2016 at 01:40:30AM +0200, Arnout Vandecappelle wrote:
[snip]
Quoted 14 lines
Quoted 14 lines
Quoted 13 lines
Quoted 4 lines
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.I think I tried running "touch" before compiling, unfortunately playing with file dates before running "make" is a bit like playing with fire. For instance with autotools based projects for which autoreconf is not run, the project may use versions of the autotools not installed on the user machine, and because of file dates may want to rerun autoconf or autmake and whine because the right versions are not available. Doing this only for .py files is much more reasonable.
Quoted 5 lines
Quoted 3 lines
I tested this as well, with make-3.81 and make-4.04. Both of them do _not_
rebuild if the timestamps are identical. And since the idea is to use touch -d @$(SOURCE_DATE_EPOCH), all timestamps will be identical.You checked both the pyo and the pyc?
Yep.
Quoted 11 lines
Quoted 3 lines
But it does indeed mean that if a package has a generated file with an earlier
date than the source files, it will now suddenly no longer be rebuilt.Yes, that is another problem. But I tried it, and this is not the one I had, the one I had was the contrary: the dates made make want to rebuild some files (the autotools/automake files), whereas the right versions of autoconf and automake were not installed, this was a package that did not run autoreconf. The "make" tool is completely based on file dates, so again, I think messing with file dates before running make is a bad idea.
Hm, I don't see how this could happen... My test showed that 'make' would not
try to rebuild the autotools/automake files when they have the same timestamp as the corresponding source file. So what triggers the reconf then?
I've tried with a random package (bash) and it's not reconfiguring after
touching all sources. Do you remember for which package you saw this?
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
PATCH 2/34 reproducibility: override locale and timezone
In reply to Thomas PetazzoniQuoted 17 lines · Thomas Petazzoni
Hello,
Quoted 3 lines · Gilles Chanteperdrix
--- Makefile | 7 +++++++ 1 file changed, 7 insertions(+)Missing SoB.
Quoted 8 lines
Makefile +0
-0@@ -218,6 +218,13 @@ ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) -include $(BR2_CONFIG) endif
Quoted 3 lines
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 221
Quoted 1 line
Makefile +1
-0+export BR2_REPRODUCIBLEWhy ?
So that it can be used in scripts. Like wrappers script. In order to know if we should take special actions because we are targeting a reproducible build.
Quoted 11 lines
Makefile ifeq ($(filter $(noconfig_targets),$(MAKECMDGOALS)),) · line 226
Quoted 6 lines
Makefile +5
-0+ifeq ($(BR2_REPRODUCIBLE),y) +export TZ=UTC +export LANG=C +export LC_ALL=C +endifRegarding overriding TZ, LANG and
LC_ALL, I'm wondering if we shouldn't do it unconditionally, regardless of the value ofBR2_REPRODUCIBLE. What is the benefit of leaving those variable to their value from the build environment?
Obviously:
- having dates printed for your local timezone (like build timestamps if you do not want a reproducible build). I do not know about you, but I would hate to have to convert a build timestamp from UTC to local timezone in order to see if I am running the binary I think I am running.
- having compiler messages printed in your locale, if you set the locale, it is probably what you want.
Signature · 2 lines
Gilles. https://click-hack.org
PATCH 9/34 reproducibility/libglib2: allow removing codegen
In reply to Arnout VandecappelleQuoted 40 lines · Arnout Vandecappelle
Quoted 3 lines · Gilles Chanteperdrix
Quoted 2 lines · Arnout Vandecappelle
Quoted 1 line · Gilles Chanteperdrix
On Tue, May 10, 2016 at 01:40:30AM +0200, Arnout Vandecappelle wrote:
[snip]
Quoted 14 lines
Quoted 14 lines
Quoted 13 lines
Quoted 4 lines
I think it would make sense to run 'touch -d @$(SOURCE_DATE_EPOCH)' on all
files after patching to capture this aspect. This was also proposed for Fedora[1] (though there it was only for the .py files). Not sure what happened with that proposal in the end.I think I tried running "touch" before compiling, unfortunately playing with file dates before running "make" is a bit like playing with fire. For instance with autotools based projects for which autoreconf is not run, the project may use versions of the autotools not installed on the user machine, and because of file dates may want to rerun autoconf or autmake and whine because the right versions are not available. Doing this only for .py files is much more reasonable.
Quoted 5 lines
Quoted 3 lines
I tested this as well, with make-3.81 and make-4.04. Both of them do _not_
rebuild if the timestamps are identical. And since the idea is to use touch -d @$(SOURCE_DATE_EPOCH), all timestamps will be identical.You checked both the pyo and the pyc?
Yep.
Quoted 11 lines
Quoted 3 lines
But it does indeed mean that if a package has a generated file with an earlier
date than the source files, it will now suddenly no longer be rebuilt.Yes, that is another problem. But I tried it, and this is not the one I had, the one I had was the contrary: the dates made make want to rebuild some files (the autotools/automake files), whereas the right versions of autoconf and automake were not installed, this was a package that did not run autoreconf. The "make" tool is completely based on file dates, so again, I think messing with file dates before running make is a bad idea.
Quoted 6 lines
Hm, I don't see how this could happen... My test showed that 'make' would not
try to rebuild the autotools/automake files when they have the same timestamp as the corresponding source file. So what triggers the reconf then?I've tried with a random package (bash) and it's not reconfiguring after
touching all sources. Do you remember for which package you saw this?
Actually, I can not remember what I did. I may not have used SOURCE_DATE_EPOCH. Anyway, I still find touching all the source files can have unintended effects, and it is not sufficient either:
- some py files may be generated by configure
- some py files may be generated by Makefiles
So, I would suggest to add a python compiler wrapper which touches the files with SOURCE_DATE_EPOCH before compiling them. That would cover all the cases and limit the unintended effects. The downside is that you have to get all packages compiling python files to use that wrapper instead of the python compiler.
Signature · 2 lines
Gilles. https://click-hack.org