[Buildroot] add option to enable AI_ADDRCONFIG support

[Buildroot] add option to enable AI_ADDRCONFIG support

[Sven Neumann]

This patchset adds a buildroot option to enable AI_ADDRCONFIG support
in uClibC. It also changes the mrouted package to select the new option
instead of hinting the user in a comment that it depends on it.

If this is accepted I will also re-submit the gupnp packages that I
tried to get into buildroot before. Until now this failed because
gssdp needs ifaddr.h which is only available with uClibC if
AI_ADDRCONFIG support is enabled.

[Buildroot] [PATCH 1/2] toolchain: add option to enable AI_ADDRCONFIG support in uclibc

[Sven Neumann]

Some packages need ifaddr.h which is only included in uclibc if
UCLIBC_SUPPORT_AI_ADDRCONFIG is selected. Make this configurable
as a buildroot option so that those packages can select it.

Signed-off-by: Sven Neumann <s.neumann@raumfeld.com>
---
 toolchain/toolchain-buildroot/Config.in.2 |    7 +++++++
 toolchain/uClibc/uclibc.mk                |    7 +++++++
 2 files changed, 14 insertions(+), 0 deletions(-)

diff --git a/toolchain/toolchain-buildroot/Config.in.2 b/toolchain/toolchain-buildroot/Config.in.2
index 7f2f68f..700d728 100644
--- a/toolchain/toolchain-buildroot/Config.in.2
+++ b/toolchain/toolchain-buildroot/Config.in.2
@@ -25,6 +25,13 @@ config BR2_TOOLCHAIN_BUILDROOT_INET_RPC
 	  Enable this option if you want your toolchain to support
 	  RPC (needed for NFS, for example).
 
+config BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG
+	bool "Enable AI ADDRCONFIG support"
+	help
+	  Enable this option if you want your toolchain to support
+	  the AI_ADDRCONFIG flag to query interfaces to find both
+	  ipv4 and ipv6 support.
+
 config BR2_TOOLCHAIN_BUILDROOT_WCHAR
 	bool "Enable WCHAR support"
 	select BR2_USE_WCHAR
diff --git a/toolchain/uClibc/uclibc.mk b/toolchain/uClibc/uclibc.mk
index ee474bb..638f918 100644
--- a/toolchain/uClibc/uclibc.mk
+++ b/toolchain/uClibc/uclibc.mk
@@ -259,6 +259,13 @@ else
 		-e 's,^.*UCLIBC_HAS_REENTRANT_RPC.*,UCLIBC_HAS_REENTRANT_RPC=n,g' \
 		$(UCLIBC_DIR)/.oldconfig
 endif
+ifeq ($(BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG),y)
+	$(SED) 's,^.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*,UCLIBC_SUPPORT_AI_ADDRCONFIG=y,g' $(UCLIBC_DIR)/.oldconfig
+else
+	$(SED) 's,.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*,UCLIBC_SUPPORT_AI_ADDRCONFIG=n,g' $(UCLIBC_DIR)/.oldconfig
+	$(SED) '/.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*/d' $(UCLIBC_DIR)/.oldconfig
+	echo "# UCLIBC_SUPPORT_AI_ADDRCONFIG is not set" >> $(UCLIBC_DIR)/.oldconfig
+endif
 ifeq ($(BR2_SOFT_FLOAT),y)
 	$(SED) 's,.*UCLIBC_HAS_FPU.*,UCLIBC_HAS_FPU=n,g' \
 		-e 's,^[^_]*HAS_FPU.*,HAS_FPU=n,g' \
-- 
1.7.5.4

[Buildroot] [PATCH 1/2] toolchain: add option to enable AI_ADDRCONFIG support in uclibc

[Peter Korsgaard]

>>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:

 Sven> Some packages need ifaddr.h which is only included in uclibc if
 Sven> UCLIBC_SUPPORT_AI_ADDRCONFIG is selected. Make this configurable
 Sven> as a buildroot option so that those packages can select it.

 Sven> +ifeq ($(BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG),y)
 Sven> +	$(SED) 's,^.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*,UCLIBC_SUPPORT_AI_ADDRCONFIG=y,g' $(UCLIBC_DIR)/.oldconfig
 Sven> +else
 Sven> +	$(SED) 's,.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*,UCLIBC_SUPPORT_AI_ADDRCONFIG=n,g' $(UCLIBC_DIR)/.oldconfig
 Sven> +	$(SED) '/.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*/d' $(UCLIBC_DIR)/.oldconfig
 Sven> +	echo "# UCLIBC_SUPPORT_AI_ADDRCONFIG is not set" >> $(UCLIBC_DIR)/.oldconfig

Why the complication and not just something like:

 $(SED) 's,^.*UCLIBC_SUPPORT_AI_ADDRCONFIG.*,#  UCLIBC_SUPPORT_AI_ADDRCONFIG is not set,g' $(UCLIBC_DIR)/.oldconfig

We'll also need to add the option for external toolchains, but I can do
that, thanks for doing the patch.

-- 
Bye, Peter Korsgaard

[Buildroot] [PATCH 2/2] mrouted: select BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG

[Sven Neumann]

Remove the comment that explains that UCLIBC_SUPPORT_AI_ADDRCONFIG
needs to be selected and select it automatically by means of the
new BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG option.

Signed-off-by: Sven Neumann <s.neumann@raumfeld.com>
---
 package/mrouted/Config.in |    4 +---
 1 files changed, 1 insertions(+), 3 deletions(-)

diff --git a/package/mrouted/Config.in b/package/mrouted/Config.in
index 1503cc6..ba68fbb 100644
--- a/package/mrouted/Config.in
+++ b/package/mrouted/Config.in
@@ -1,9 +1,7 @@
 config BR2_PACKAGE_MROUTED
 	bool "mrouted"
+	select BR2_TOOLCHAIN_BUILDROOT_INET_ADDRCONFIG
 	help
 	  An implementation of the DVMRP multicast routing protocol.
 
-	  If built with uClibc, then UCLIBC_SUPPORT_AI_ADDRCONFIG is
-	  needed in uClibc configuration.
-
 	  http://github.com/troglobit/mrouted
-- 
1.7.5.4

[Buildroot] add option to enable AI_ADDRCONFIG support

[Thomas Petazzoni]

Hello Sven,

Le Wed,  7 Mar 2012 23:38:16 +0100,
Sven Neumann <s.neumann@raumfeld.com> a ?crit :

> This patchset adds a buildroot option to enable AI_ADDRCONFIG support
> in uClibC. It also changes the mrouted package to select the new option
> instead of hinting the user in a comment that it depends on it.

I don't know if this question has been asked in the past, but do you
know what is the binary size impact of enabling AI_ADDRCONFIG in
uClibc? If the impact is rather minimal, then I'd rather have it
enabled in our default configuration instead of adding yet another
toolchain config knob.

Anyway some comments about your patches (but don't work on updating
them before we decide if a new option is really needed) :

 * Your PATCH 1/2 also needs to address the cases of external
   toolchains and crosstool-ng toolchains ;

 * Your PATCH 2/2 should use a depends on rather than a select. We use
   depends on for all the dependencies of packages on toolchain
   options, together with a comment displayed when the toolchain
   options isn't present, so that the user is aware that mrouted is
   available, but only if AI_ADDRCONFIG is enabled as a toolchain
   option.

Regards,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] add option to enable AI_ADDRCONFIG support

[Peter Korsgaard]

>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes:

Hi,

 Thomas> I don't know if this question has been asked in the past, but do you
 Thomas> know what is the binary size impact of enabling AI_ADDRCONFIG in
 Thomas> uClibc? If the impact is rather minimal, then I'd rather have it
 Thomas> enabled in our default configuration instead of adding yet another
 Thomas> toolchain config knob.

I believe it's a fairly small impact, but the problem is external uClibc
toolchains (made by BR, ct-NG or elsewhere) which most likely won't have
AI_ADDRCONFIG enabled, so mrouted (and gupnp) will fail at build time.


 Thomas> Anyway some comments about your patches (but don't work on updating
 Thomas> them before we decide if a new option is really needed) :

 Thomas>  * Your PATCH 1/2 also needs to address the cases of external
 Thomas>    toolchains and crosstool-ng toolchains ;

 Thomas>  * Your PATCH 2/2 should use a depends on rather than a select. We use
 Thomas>    depends on for all the dependencies of packages on toolchain
 Thomas>    options, together with a comment displayed when the toolchain
 Thomas>    options isn't present, so that the user is aware that mrouted is
 Thomas>    available, but only if AI_ADDRCONFIG is enabled as a toolchain
 Thomas>    option.

I agree.

-- 
Bye, Peter Korsgaard

[Buildroot] add option to enable AI_ADDRCONFIG support

[Thomas Petazzoni]

Le Sun, 11 Mar 2012 21:56:22 +0100,
Peter Korsgaard <jacmet@uclibc.org> a ?crit :

> I believe it's a fairly small impact, but the problem is external uClibc
> toolchains (made by BR, ct-NG or elsewhere) which most likely won't have
> AI_ADDRCONFIG enabled, so mrouted (and gupnp) will fail at build time.

Is it common to have AI_ADDRCONFIG disabled in external toolchains?
Notice that we have a similar problem with ltp-testsuite at the moment.

Do we really want to add one more toolchain option just for this small
feature? Sounds like a big hammer, no?

Regards,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] add option to enable AI_ADDRCONFIG support

[Peter Korsgaard]

>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes:

Hi,

 >> I believe it's a fairly small impact, but the problem is external uClibc
 >> toolchains (made by BR, ct-NG or elsewhere) which most likely won't have
 >> AI_ADDRCONFIG enabled, so mrouted (and gupnp) will fail at build time.

 Thomas> Is it common to have AI_ADDRCONFIG disabled in external toolchains?
 Thomas> Notice that we have a similar problem with ltp-testsuite at the moment.

Hard to say, but the config option defaults to 'n' in uClibc - So
chances are that it is.

 Thomas> Do we really want to add one more toolchain option just for
 Thomas> this small feature? Sounds like a big hammer, no?

True, but it is the safest approach. I don't feel strongly about it
though - Anybody?

-- 
Bye, Peter Korsgaard

[Buildroot] add option to enable AI_ADDRCONFIG support

[Gustavo Zacarias]

On 2012-03-15 17:56, Peter Korsgaard wrote:

>  Thomas> Is it common to have AI_ADDRCONFIG disabled in external 
> toolchains?
>  Thomas> Notice that we have a similar problem with ltp-testsuite at
> the moment.
>
> Hard to say, but the config option defaults to 'n' in uClibc - So
> chances are that it is.
>
>  Thomas> Do we really want to add one more toolchain option just for
>  Thomas> this small feature? Sounds like a big hammer, no?
>
> True, but it is the safest approach. I don't feel strongly about it
> though - Anybody?

IMHO it should be enabled by default, it's been in glibc since version 
2.3.
Some programs benefit a lot in the configuration department from it, 
for example samba.
Doing a interfaces=192.168.1.1/255/255/255.0 statement versus a 
interfaces=eth0 statement in the configuration file is a big bonus to 
me.
I'm sending a patch right away.
Regards.