Skip to contentSkip to conversation

Strip binaries in the fakeroot script instead of in target-finalize

PATCH New unixmania at gmail.com 2 participants 3 messages Sep 30, 2019 – Oct 1, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Strip binaries in the fakeroot script instead of in target-finalize

From: Carlos Santos <[email protected]>

Since commit 118534fe54 the root filesystem image is generated from an intermediary copy of TARGET_DIR, so we can move the stripping of target binaries from target-finalize to the fakeroot script.

This allows us to easily find the non-stripped executables to debug with gdbserver, as they are at the same relative path in TARGET_DIR as in the target device, rather than searching inside the build directory.

We can also set sysroot to TARGET_DIR in gdb, instead of STAGING_DIR.

---
CC: Ciro Santilli <[email protected]>
---

---
 Makefile     | 34 ----------------------------------
 fs/common.mk | 34 ++++++++++++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 34 deletions(-)
Makefile +0 -34
index ecaae13846..57314e5f1d 100644
@@ -614,38 +614,6 @@ RSYNC_VCS_EXCLUSIONS = \
 	--exclude .svn --exclude .git --exclude .hg --exclude .bzr \
 	--exclude CVS
 
-# When stripping, obey to BR2_STRIP_EXCLUDE_DIRS and
-# BR2_STRIP_EXCLUDE_FILES
-STRIP_FIND_COMMON_CMD = \
-	find $(TARGET_DIR) \
-	$(if $(call qstrip,$(BR2_STRIP_EXCLUDE_DIRS)), \
-		\( $(call finddirclauses,$(TARGET_DIR),$(call qstrip,$(BR2_STRIP_EXCLUDE_DIRS))) \) \
-		-prune -o \
-	) \
-	$(if $(call qstrip,$(BR2_STRIP_EXCLUDE_FILES)), \
-		-not \( $(call findfileclauses,$(call qstrip,$(BR2_STRIP_EXCLUDE_FILES))) \) )
-
-# Regular stripping for everything, except libpthread, ld-*.so and
-# kernel modules:
-# - libpthread.so: a non-stripped libpthread shared library is needed for
-#   proper debugging of pthread programs using gdb.
-# - ld.so: a non-stripped dynamic linker library is needed for valgrind
-# - kernel modules (*.ko): do not function properly when stripped like normal
-#   applications and libraries. Normally kernel modules are already excluded
-#   by the executable permission check, so the explicit exclusion is only
-#   done for kernel modules with incorrect permissions.
-STRIP_FIND_CMD = \
-	$(STRIP_FIND_COMMON_CMD) \
-	-type f \( -perm /111 -o -name '*.so*' \) \
-	-not \( $(call findfileclauses,libpthread*.so* ld-*.so* *.ko) \) \
-	-print0
-
-# Special stripping (only debugging symbols) for libpthread and ld-*.so.
-STRIP_FIND_SPECIAL_LIBS_CMD = \
-	$(STRIP_FIND_COMMON_CMD) \
-	\( -name 'ld-*.so*' -o -name 'libpthread*.so*' \) \
-	-print0
-
 ifeq ($(BR2_ECLIPSE_REGISTER),y)
 define TOOLCHAIN_ECLIPSE_REGISTER
 	./support/scripts/eclipse-register-toolchain `readlink -f $(O)` \
@@ -761,8 +729,6 @@ endif
 	rm -rf $(TARGET_DIR)/usr/doc $(TARGET_DIR)/usr/share/doc
 	rm -rf $(TARGET_DIR)/usr/share/gtk-doc
 	rmdir $(TARGET_DIR)/usr/share 2>/dev/null || true
-	$(STRIP_FIND_CMD) | xargs -0 $(STRIPCMD) 2>/dev/null || true
-	$(STRIP_FIND_SPECIAL_LIBS_CMD) | xargs -0 -r $(STRIPCMD) $(STRIP_STRIP_DEBUG) 2>/dev/null || true
 
 	test -f $(TARGET_DIR)/etc/ld.so.conf && \
 		{ echo "ERROR: we shouldn't have a /etc/ld.so.conf file"; exit 1; } || true
fs/common.mk +34 -0
index 842ea924a5..d835d7dccc 100644
@@ -61,6 +61,38 @@ ROOTFS_COMMON_FINAL_RECURSIVE_DEPENDENCIES = $(sort \
 	) \
 	$(ROOTFS_COMMON_FINAL_RECURSIVE_DEPENDENCIES__X))
 
+# When stripping, obey to BR2_STRIP_EXCLUDE_DIRS and
+# BR2_STRIP_EXCLUDE_FILES
+STRIP_FIND_COMMON_CMD = \
+	find $(TARGET_DIR) \
+	$(if $(call qstrip,$(BR2_STRIP_EXCLUDE_DIRS)), \
+		\( $(call finddirclauses,$(TARGET_DIR),$(call qstrip,$(BR2_STRIP_EXCLUDE_DIRS))) \) \
+		-prune -o \
+	) \
+	$(if $(call qstrip,$(BR2_STRIP_EXCLUDE_FILES)), \
+		-not \( $(call findfileclauses,$(call qstrip,$(BR2_STRIP_EXCLUDE_FILES))) \) )
+
+# Regular stripping for everything, except libpthread, ld-*.so and
+# kernel modules:
+# - libpthread.so: a non-stripped libpthread shared library is needed for
+#   proper debugging of pthread programs using gdb.
+# - ld.so: a non-stripped dynamic linker library is needed for valgrind
+# - kernel modules (*.ko): do not function properly when stripped like normal
+#   applications and libraries. Normally kernel modules are already excluded
+#   by the executable permission check, so the explicit exclusion is only
+#   done for kernel modules with incorrect permissions.
+STRIP_FIND_CMD = \
+	$(STRIP_FIND_COMMON_CMD) \
+	-type f \( -perm /111 -o -name '*.so*' \) \
+	-not \( $(call findfileclauses,libpthread*.so* ld-*.so* *.ko) \) \
+	-print0
+
+# Special stripping (only debugging symbols) for libpthread and ld-*.so.
+STRIP_FIND_SPECIAL_LIBS_CMD = \
+	$(STRIP_FIND_COMMON_CMD) \
+	\( -name 'ld-*.so*' -o -name 'libpthread*.so*' \) \
+	-print0
+
 .PHONY: rootfs-common
 rootfs-common: $(ROOTFS_COMMON_DEPENDENCIES) target-finalize
 	@$(call MESSAGE,"Generating root filesystems common tables")
@@ -160,6 +192,8 @@ $$(BINARIES_DIR)/$$(ROOTFS_$(2)_FINAL_IMAGE_NAME): $$(ROOTFS_$(2)_DEPENDENCIES)
 	echo '#!/bin/sh' > $$(FAKEROOT_SCRIPT)
 	echo "set -e" >> $$(FAKEROOT_SCRIPT)
 
+	echo "$$(STRIP_FIND_CMD) | xargs -0 $$(STRIPCMD) 2>/dev/null || true" >> $$(FAKEROOT_SCRIPT)
+	echo "$$(STRIP_FIND_SPECIAL_LIBS_CMD) | xargs -0 -r $$(STRIPCMD) $$(STRIP_STRIP_DEBUG) 2>/dev/null || true" >> $$(FAKEROOT_SCRIPT)
 	echo "chown -h -R 0:0 $$(TARGET_DIR)" >> $$(FAKEROOT_SCRIPT)
 	PATH=$$(BR_PATH) $$(TOPDIR)/support/scripts/mkusers $$(ROOTFS_FULL_USERS_TABLE) $$(TARGET_DIR) >> $$(FAKEROOT_SCRIPT)
 	echo "$$(HOST_DIR)/bin/makedevs -d $$(ROOTFS_FULL_DEVICES_TABLE) $$(TARGET_DIR)" >> $$(FAKEROOT_SCRIPT)
Signature · 1 line
2.18.1
#1 lore raw

PATCH 1/1 Strip binaries in the fakeroot script instead of in target-finalize

Quoted 9 lines · unixmania at gmail.com

From: Carlos Santos <[email protected]>

Since commit 118534fe54 the root filesystem image is generated from an intermediary copy of TARGET_DIR, so we can move the stripping of target binaries from target-finalize to the fakeroot script.

This allows us to easily find the non-stripped executables to debug with gdbserver, as they are at the same relative path in TARGET_DIR as in the target device, rather than searching inside the build directory.

This is in fact a pretty elegant solution, even compared to the
target-is-a-subset-of-staging approach that we'd like to move to long-term. Indeed, anything which is not in target should never be relevant for the debugger. E.g. .a files are not needed for debugging, and header files will anyway be searched in the full STAGING_DIR path.

It's a bit unelegant to do this in the fakeroot script, though, because it has
nothing to do with fakeroot. However, it can just as well be done a few lines higher, after the rsync and before the fakeroot script is generated.

So, with that changed, you can add my

Still, I'd like to have Yann take a look as well.

Quoted 1 line

We can also set sysroot to TARGET_DIR in gdb, instead of STAGING_DIR.

So, will you send a second patch that does exactly that [1] ?

Regards,
Arnout

[1]
https://git.buildroot.org/buildroot/tree/toolchain/helpers.mk?id=b3f033282bbada2e85625583fbcf56db0031d19e#n462

Quoted 6 lines

---
CC: Ciro Santilli <[email protected]>

#2 lore raw

PATCH 1/1 Strip binaries in the fakeroot script instead of in target-finalize

Quoted 20 lines · Arnout Vandecappelle
Quoted 9 lines · unixmania at gmail.com

From: Carlos Santos <[email protected]>

Since commit 118534fe54 the root filesystem image is generated from an intermediary copy of TARGET_DIR, so we can move the stripping of target binaries from target-finalize to the fakeroot script.

This allows us to easily find the non-stripped executables to debug with gdbserver, as they are at the same relative path in TARGET_DIR as in the target device, rather than searching inside the build directory.

This is in fact a pretty elegant solution, even compared to the
target-is-a-subset-of-staging approach that we'd like to move to long-term. Indeed, anything which is not in target should never be relevant for the debugger. E.g. .a files are not needed for debugging, and header files will anyway be searched in the full STAGING_DIR path.

Quoted 3 lines

It's a bit unelegant to do this in the fakeroot script, though, because it has
nothing to do with fakeroot. However, it can just as well be done a few lines higher, after the rsync and before the fakeroot script is generated.

Good idea.

Quoted 11 lines

So, with that changed, you can add my

Still, I'd like to have Yann take a look as well.

Quoted 1 line

We can also set sysroot to TARGET_DIR in gdb, instead of STAGING_DIR.

So, will you send a second patch that does exactly that [1] ?

Yes

Quoted 13 lines

Regards,
Arnout

[1]
https://git.buildroot.org/buildroot/tree/toolchain/helpers.mk?id=b3f033282bbada2e85625583fbcf56db0031d19e#n462

Quoted 6 lines

---
CC: Ciro Santilli <[email protected]>

Signature · 1 line
Carlos Santos <[email protected]>