Skip to contentSkip to conversation

package/dbus: fix group and mode for the launcher

PATCH New Norbert Lange 3 participants 8 messages Jun 5, 2020 – Jul 12, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 package/dbus: fix group and mode for the launcher

see bus/Makefile.am

---
 package/dbus/dbus.mk | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
package/dbus/dbus.mk +1 -1
index a4043864d1..bb9f17a5e0 100644
@@ -11,7 +11,7 @@ DBUS_LICENSE_FILES = COPYING
 DBUS_INSTALL_STAGING = YES
 
 define DBUS_PERMISSIONS
-	/usr/libexec/dbus-daemon-launch-helper f 4755 0 0 - - - - -
+	/usr/libexec/dbus-daemon-launch-helper f 4750 0 dbus - - - - -
 endef
 
 define DBUS_USERS
Signature · 1 line
2.26.2
#1 lore raw

PATCH 2/2 package/dbus: remove /usr/lib/dbus-1.0 if empty

---
 package/dbus/dbus.mk | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
package/dbus/dbus.mk +2 -1
index bb9f17a5e0..7a7762f537 100644
@@ -80,7 +80,8 @@ endef
 DBUS_PRE_INSTALL_TARGET_HOOKS += DBUS_REMOVE_VAR_LIB_DBUS
 
 define DBUS_REMOVE_DEVFILES
-	rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0
+	rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0/include
+	-rmdir $(TARGET_DIR)/usr/lib/dbus-1.0 2>/dev/null
 endef
 
 DBUS_POST_INSTALL_TARGET_HOOKS += DBUS_REMOVE_DEVFILES
Signature · 1 line
2.26.2
#2 lore raw

PATCH 2/2 package/dbus: remove /usr/lib/dbus-1.0 if empty

Quoted 3 lines · On Sat, 6 Jun 2020 00:52:46 +0200 Norbert Lange <[email protected]> wrote:
---
 package/dbus/dbus.mk | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Your Signed-off-by is missing, and there is no commit log. We clearly need a better explanation. And your commit title is wrong: you're not just removing /usr/lib/dbus-1.0 is empty: you're changing from removing it unconditionally to removing just /usr/lib/dbus-1.0/include and then removing /usr/lib/dbus-1.0 if it's empty.

So it means that you have installed *something* in /usr/lib/dbus-1.0 that you want Buildroot to keep on the target. But what? This is not explained in your commit log.

package/dbus/dbus.mk endef · line 84

Quoted 11 lines
package/dbus/dbus.mk +2 -1
index bb9f17a5e0..7a7762f537 100644
@@ -80,7 +80,8 @@ endef
 DBUS_PRE_INSTALL_TARGET_HOOKS += DBUS_REMOVE_VAR_LIB_DBUS
 
 define DBUS_REMOVE_DEVFILES
-	rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0
+	rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0/include
+	-rmdir $(TARGET_DIR)/usr/lib/dbus-1.0 2>/dev/null

Can we use --ignore-fail-on-non-empty instead ? Yann, is this option reasonably available even on old distributions ?

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#3 lore raw

PATCH 2/2 package/dbus: remove /usr/lib/dbus-1.0 if empty

Am Sa., 6. Juni 2020 um 22:38 Uhr schrieb Thomas Petazzoni <[email protected]>:

Quoted 8 lines
Quoted 3 lines · On Sat, 6 Jun 2020 00:52:46 +0200 Norbert Lange <[email protected]> wrote:
---
 package/dbus/dbus.mk | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 5 lines

Your Signed-off-by is missing, and there is no commit log. We clearly need a better explanation. And your commit title is wrong: you're not just removing /usr/lib/dbus-1.0 is empty: you're changing from removing it unconditionally to removing just /usr/lib/dbus-1.0/include and then removing /usr/lib/dbus-1.0 if it's empty.

Ok, gonna improve that in a v2 if you find no other objections

Quoted 4 lines

So it means that you have installed *something* in /usr/lib/dbus-1.0 that you want Buildroot to keep on the target. But what? This is not explained in your commit log.

I wrote this commit some months ago, would have to outright lie, but I think it was dbusbroker (patch on the ML).
The basic gist is, that /usr/lib/dbus-1.0 is some sort of libexec dir (on Debian and Fedora AFAIK), and I believe dbusbroker installs it stuff there.

What we really want is to remove the unneeded files (the includes), and the parent directory it is empty.

Quoted 16 lines

package/dbus/dbus.mk endef · line 84

Quoted 12 lines
package/dbus/dbus.mk +2 -1
index bb9f17a5e0..7a7762f537 100644
@@ -80,7 +80,8 @@ endef
 DBUS_PRE_INSTALL_TARGET_HOOKS += DBUS_REMOVE_VAR_LIB_DBUS
 
 define DBUS_REMOVE_DEVFILES
-     rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0
+     rm -rf $(TARGET_DIR)/usr/lib/dbus-1.0/include
+     -rmdir $(TARGET_DIR)/usr/lib/dbus-1.0 2>/dev/null

Can we use --ignore-fail-on-non-empty instead ? Yann, is this option reasonably available even on old distributions ?

Quoted 5 lines

Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Norbert

#4 lore raw

PATCH 2/2 package/dbus: remove /usr/lib/dbus-1.0 if empty

Hello Norbert,

Quoted 7 lines · On Sun, 7 Jun 2020 00:36:47 +0200 Norbert Lange <[email protected]> wrote:
Quoted 5 lines

Your Signed-off-by is missing, and there is no commit log. We clearly need a better explanation. And your commit title is wrong: you're not just removing /usr/lib/dbus-1.0 is empty: you're changing from removing it unconditionally to removing just /usr/lib/dbus-1.0/include and then removing /usr/lib/dbus-1.0 if it's empty.

Ok, gonna improve that in a v2 if you find no other objections

My main objection is that we need a better commit log.

Quoted 4 lines
Quoted 3 lines

So it means that you have installed *something* in /usr/lib/dbus-1.0 that you want Buildroot to keep on the target. But what? This is not explained in your commit log.

Quoted 8 lines

I wrote this commit some months ago, would have to outright lie, but I think it was dbusbroker (patch on the ML).
The basic gist is, that /usr/lib/dbus-1.0 is some sort of libexec dir (on Debian and Fedora AFAIK), and I believe dbusbroker installs it stuff there.

What we really want is to remove the unneeded files (the includes), and the parent directory it is empty.

Could you verify what dbusbroker installs in this directory, explain why it is relevant, and use that as a justification in the commit log ?

Thanks a lot!

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#5 lore raw

PATCH 2/2 package/dbus: remove /usr/lib/dbus-1.0 if empty

Am So., 7. Juni 2020 um 10:45 Uhr schrieb Thomas Petazzoni <[email protected]>:

Quoted 32 lines

Hello Norbert,

Quoted 7 lines · On Sun, 7 Jun 2020 00:36:47 +0200 Norbert Lange <[email protected]> wrote:
Quoted 5 lines

Your Signed-off-by is missing, and there is no commit log. We clearly need a better explanation. And your commit title is wrong: you're not just removing /usr/lib/dbus-1.0 is empty: you're changing from removing it unconditionally to removing just /usr/lib/dbus-1.0/include and then removing /usr/lib/dbus-1.0 if it's empty.

Ok, gonna improve that in a v2 if you find no other objections

My main objection is that we need a better commit log.

Quoted 4 lines
Quoted 3 lines

So it means that you have installed *something* in /usr/lib/dbus-1.0 that you want Buildroot to keep on the target. But what? This is not explained in your commit log.

Quoted 8 lines

I wrote this commit some months ago, would have to outright lie, but I think it was dbusbroker (patch on the ML).
The basic gist is, that /usr/lib/dbus-1.0 is some sort of libexec dir (on Debian and Fedora AFAIK), and I believe dbusbroker installs it stuff there.

What we really want is to remove the unneeded files (the includes), and the parent directory it is empty.

Could you verify what dbusbroker installs in this directory, explain why it is relevant, and use that as a justification in the commit log ?

Quoted 7 lines

Thanks a lot!

Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Well,

turns out I was playing with libexec configuration, and dbus then places its helpers in /usr/lib/dbus-1.0.
So, it's actually useless for the upstream buildroot (as you wont allow for libexec config), but it actually wont hurt either. Amended patch is following.

#6 lore raw

PATCH 1/2 package/dbus: fix group and mode for the launcher

In reply to Norbert Lange
Quoted 3 lines · On Sat, 6 Jun 2020 00:52:45 +0200 Norbert Lange <[email protected]> wrote:

see bus/Makefile.am

This commit log was really too short in details. I've expanded it with more details and applied.

Thanks!

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com