busybox: Include setsid and cttyhack in defconfig

When building very small systems, it can be useful to spawn a shell from a simple init script, rather than a full System V Init process. This requires the shell be the session leader and be able to open the controlling terminal if it is to have job control.

Enable CONFIG_CTTYHACK and CONFIG_SETSID to enable this. The resulting busybox binary does not change in size (I checked several times!).

Signed-off-by Darren Hart <[email protected]>

---
 meta/recipes-core/busybox/busybox-1.19.4/defconfig |    4 ++--
 meta/recipes-core/busybox/busybox_1.19.4.bb        |    2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

meta/recipes-core/busybox/busybox-1.19.4/defconfig +2 -2

index 372d7b5..0a5d5c9 100644
@@ -700,7 +700,7 @@ CONFIG_MICROCOM=y
 # CONFIG_RFKILL is not set
 # CONFIG_RUNLEVEL is not set
 # CONFIG_RX is not set
-# CONFIG_SETSID is not set
+CONFIG_SETSID=y
 CONFIG_STRINGS=y
 # CONFIG_TASKSET is not set
 # CONFIG_FEATURE_TASKSET_FANCY is not set

@@ -960,7 +960,7 @@ CONFIG_ASH_BUILTIN_TEST=y
 CONFIG_ASH_OPTIMIZE_FOR_SIZE=y
 # CONFIG_ASH_RANDOM_SUPPORT is not set
 CONFIG_ASH_EXPAND_PRMT=y
-# CONFIG_CTTYHACK is not set
+CONFIG_CTTYHACK=y
 # CONFIG_HUSH is not set
 # CONFIG_HUSH_BASH_COMPAT is not set
 # CONFIG_HUSH_BRACE_EXPANSION is not set

meta/recipes-core/busybox/busybox_1.19.4.bb +1 -1

index 0717075..5b3f356 100644
@@ -1,5 +1,5 @@
 require busybox.inc
-PR = "r6"
+PR = "r7"
 
 SRC_URI = "http://www.busybox.net/downloads/busybox-${PV}.tar.bz2;name=tarball \
            file://B921600.patch \

So the delta for including SETSID and CTTYHACK is 2560 bytes.

Quoted 1 line · Darren Hart

So the delta for including SETSID and CTTYHACK is 2560 bytes.

Personally I am still not in favour of adding this to the default configuration. I appreciate that it's "only" 2.5k in this case, but every time we make a change like this the binary gets a little bit bigger and, over time, it does add up. This sort of gradual bloat is quite insidious and difficult to combat after the fact.

So, I continue to think that poky-tiny should just provide its own busybox configuration and turn on the options that it wants just like other DISTROs do. No doubt there are some things currently included in the oe-core defaults that poky-tiny doesn't need, so you would probably get a smaller binary that way as well.

Quoted 9 lines · Phil Blundell

Quoted 1 line · Darren Hart

So the delta for including SETSID and CTTYHACK is 2560 bytes.

Personally I am still not in favour of adding this to the default configuration.  I appreciate that it's "only" 2.5k in this case, but every time we make a change like this the binary gets a little bit bigger and, over time, it does add up.  This sort of gradual bloat is quite insidious and difficult to combat after the fact.

Quoted 5 lines

So, I continue to think that poky-tiny should just provide its own busybox configuration and turn on the options that it wants just like other DISTROs do.  No doubt there are some things currently included in the oe-core defaults that poky-tiny doesn't need, so you would probably get a smaller binary that way as well.

You are correct that poky-tiny would benefit from a smaller config. My original intent was to update the busybox recipe to use the new merge-config.sh that we pushed to the upstream Linux kernel (which should work with busybox as it uses the same config mechanism). This would allow us to maintain a base busybox config with a several config fragments that can be easily added via DISTRO_FEATURES rather than the complicated hack that is in busybox now for handling DISTRO_FEATURES. I prefer this approach as it reduces (if not eliminates) the need for the proliferation of busybox.bbappend files.

However, this is a larger project and my immediate goal is to get poky-tiny into better shape in terms of the initial experience. This is why I originally implemented it as a "tiny" DISTRO_FEATURE as that would migrate naturally to the config fragment approach. You and others objected to that approach, and I do understand not wanting to complicate the DISTRO_FEATURE logic further.

With the above goal in mind, can you accept either of my proposed patches as an interim solution? Either as an added tiny DISTRO_FEATURE or as a simple addition to the defconfig? I do believe these two features are useful beyond poky-tiny.

Quoted 12 lines · Darren Hart

Quoted 9 lines · Phil Blundell

Quoted 1 line · Darren Hart

So the delta for including SETSID and CTTYHACK is 2560 bytes.

Personally I am still not in favour of adding this to the default configuration. I appreciate that it's "only" 2.5k in this case, but every time we make a change like this the binary gets a little bit bigger and, over time, it does add up. This sort of gradual bloat is quite insidious and difficult to combat after the fact.

Quoted 5 lines

Quoted 5 lines

So, I continue to think that poky-tiny should just provide its own busybox configuration and turn on the options that it wants just like other DISTROs do. No doubt there are some things currently included in the oe-core defaults that poky-tiny doesn't need, so you would probably get a smaller binary that way as well.

in retrospect I agree with Phil on gradual bloat. busybox and other kconfig based
packages will always have fine tunings that we can never say one size fits all unless
you enable everything and I think the purpose of using kconfig in those packages is to provide this fine level of configuration

Quoted 11 lines

You are correct that poky-tiny would benefit from a smaller config. My original intent was to update the busybox recipe to use the new merge-config.sh that we pushed to the upstream Linux kernel (which should work with busybox as it uses the same config mechanism). This would allow us to maintain a base busybox config with a several config fragments that can be easily added via DISTRO_FEATURES rather than the complicated hack that is in busybox now for handling DISTRO_FEATURES. I prefer this approach as it reduces (if not eliminates) the need for the proliferation of busybox.bbappend files.

. using merge-config.sh will probably make things better but until then I don't think its a bad thing to have bbappends in current scenario

Quoted 7 lines

However, this is a larger project and my immediate goal is to get poky-tiny into better shape in terms of the initial experience. This is why I originally implemented it as a "tiny" DISTRO_FEATURE as that would migrate naturally to the config fragment approach. You and others objected to that approach, and I do understand not wanting to complicate the DISTRO_FEATURE logic further.

Quoted 4 lines

With the above goal in mind, can you accept either of my proposed patches as an interim solution? Either as an added tiny DISTRO_FEATURE or as a simple addition to the defconfig? I do believe these two features are useful beyond poky-tiny.

I think best approach here is to have the defconfig of own in poky-tiny layer. It will
be a contained change.

Quoted 54 lines · Darren Hart

Quoted 12 lines · Phil Blundell

Quoted 9 lines · Darren Hart

Quoted 1 line

So the delta for including SETSID and CTTYHACK is 2560 bytes.

Personally I am still not in favour of adding this to the default configuration. I appreciate that it's "only" 2.5k in this case, but every time we make a change like this the binary gets a little bit bigger and, over time, it does add up. This sort of gradual bloat is quite insidious and difficult to combat after the fact.

Quoted 5 lines

Quoted 5 lines

So, I continue to think that poky-tiny should just provide its own busybox configuration and turn on the options that it wants just like other DISTROs do. No doubt there are some things currently included in the oe-core defaults that poky-tiny doesn't need, so you would probably get a smaller binary that way as well.

in retrospect I agree with Phil on gradual bloat. busybox and other kconfig based
packages will always have fine tunings that we can never say one size fits all unless
you enable everything and I think the purpose of using kconfig in those packages is to provide this fine level of configuration

Quoted 11 lines

You are correct that poky-tiny would benefit from a smaller config. My original intent was to update the busybox recipe to use the new merge-config.sh that we pushed to the upstream Linux kernel (which should work with busybox as it uses the same config mechanism). This would allow us to maintain a base busybox config with a several config fragments that can be easily added via DISTRO_FEATURES rather than the complicated hack that is in busybox now for handling DISTRO_FEATURES. I prefer this approach as it reduces (if not eliminates) the need for the proliferation of busybox.bbappend files.

. using merge-config.sh will probably make things better but until then I don't think its a bad thing to have bbappends in current scenario

Quoted 11 lines

However, this is a larger project and my immediate goal is to get poky-tiny into better shape in terms of the initial experience. This is why I originally implemented it as a "tiny" DISTRO_FEATURE as that would migrate naturally to the config fragment approach. You and others objected to that approach, and I do understand not wanting to complicate the DISTRO_FEATURE logic further.

With the above goal in mind, can you accept either of my proposed patches as an interim solution? Either as an added tiny DISTRO_FEATURE or as a simple addition to the defconfig? I do believe these two features are useful beyond poky-tiny.

Quoted 3 lines

I think best approach here is to have the defconfig of own in poky-tiny layer. It will
be a contained change.

Note that poky-tiny is not currently in its own layer, but is included in the meta-intel layer. I will look into ways to accomplish this without adding to the core busybox config and not having to break out poky-tiny into its own layer.

Op 16 jun. 2012, om 16:47 heeft Darren Hart het volgende geschreven:

Quoted 4 lines

Note that poky-tiny is not currently in its own layer, but is included in the meta-intel layer. I will look into ways to accomplish this without adding to the core busybox config and not having to break out poky-tiny into its own layer.

What is it with you intel folks (bless their hearts) wanting to combine BSP and distro? baerion, guacamayo and poky-tiny all force you to include the distro stuff if you want any of their recipes. I don't want to have all the hostnames on my boards changed to 'baerion' when I build mediatomb!

Firstly, I think when Darren said "meta-intel" he meant to say "meta-yocto" which is where the poky-tiny distro config can be found (and indeed that is where I would expect to find it).

Quoted 10 lines

Op 16 jun. 2012, om 16:47 heeft Darren Hart het volgende geschreven:

Quoted 4 lines

Note that poky-tiny is not currently in its own layer, but is included in the meta-intel layer. I will look into ways to accomplish this without adding to the core busybox config and not having to break out poky-tiny into its own layer.

What is it with you intel folks (bless their hearts) wanting to combine BSP and distro? baerion, guacamayo and poky-tiny all force you to include the distro stuff if you want any of their recipes. I don't want to have all the hostnames on my boards changed to 'baerion' when I build mediatomb!

All of these efforts are separate and put together by different people.

Firstly, I think when Darren said "meta-intel" he meant to say "meta-yocto" which is where the poky-tiny distro config can be found (and indeed that is where I would expect to find it).

Secondly, you already asked me about Baryon and I already agreed that it needed reorganising and promised to do it; but I haven't had time yet. I'd also point out in any case that it does not contain any BSP and the mere presence of the baryon distro config does not force you to use it as you suggest.

Paul

Signature · 2 lines

Paul Eggleton
Intel Open Source Technology Centre

it does force the distro bbappends onto everyone, messing up /etc/hostname

Quoted 1 line

it does force the distro bbappends onto everyone, messing up /etc/hostname

So it did. Now it's fixed.

Op 18 jun. 2012, om 09:58 heeft Paul Eggleton het volgende geschreven:

Quoted 5 lines · On Monday 18 June 2012 07:00:19 Koen Kooi wrote:

Quoted 2 lines

it does force the distro bbappends onto everyone, messing up /etc/hostname

So it did. Now it's fixed.

Thanks! I'll add it to the angstrom layer stack later this week.

With the above goal in mind, can you accept either of my proposed patches as an interim solution? Either as an added tiny DISTRO_FEATURE or as a simple addition to the defconfig? I do believe these two features are useful beyond poky-tiny.

Quoted 4 lines · Darren Hart

With the above goal in mind, can you accept either of my proposed patches as an interim solution? Either as an added tiny DISTRO_FEATURE or as a simple addition to the defconfig? I do believe these two features are useful beyond poky-tiny.

Well, obviously it isn't my place to accept or reject your patch; one of the oe-core maintainers would have to make that decision. But, personally, given a choice between those two options I would prefer the latter.

However, I do still think that the best interim approach is just to provide a suitable, complete defconfig file in whatever layer contains poky-tiny. This doesn't need to involve splitting poky-tiny into its own layer, doesn't require any extra tooling in oe-core in the short term, and doesn't necessarily even need to involve a .bbappend (although it would probably be easiest if it did).