[v4] rfkill: new package

Add rfkill package.

Signed-off-by: Viacheslav Volkov <sv99@inbox.ru>
---
Changes v3 -> v4:
  - remove trailing space in the comment  (suggested by yann.morin.1998@free.fr)
  - remove trailing slash in the URL

Changes v2 -> v3:
  - wrapp comment to 72 characters length  (suggested by Vincent.Riera@imgtec.com)

Changes v1 -> v2:
  - correct Config.in  (suggested by Vincent.Riera@imgtec.com)
  - correct rfkill.mk
---
 package/Config.in          |  1 +
 package/rfkill/Config.in   |  7 +++++++
 package/rfkill/rfkill.hash |  2 ++
 package/rfkill/rfkill.mk   | 20 ++++++++++++++++++++
 4 files changed, 30 insertions(+)
 create mode 100644 package/rfkill/Config.in
 create mode 100644 package/rfkill/rfkill.hash
 create mode 100644 package/rfkill/rfkill.mk

Viacheslav, All,

I said in my previous review that you should wait a bit before
resending, in case other may have mor ecomments.

When I said wait a bit, I really meant: wait until tomorrow (at the very
least).

Since the previous review, we discussed on IRC, and we discovered
more issues with your patch, see below...

Be sure to read up until the end of the reply, there ar emore than one
comment to it...

On 2015-10-27 18:32 +0300, Viacheslav Volkov spake thusly:
> Add rfkill package.
> 
> Signed-off-by: Viacheslav Volkov <sv99@inbox.ru>
[--SNIP--]
> diff --git a/package/rfkill/rfkill.hash b/package/rfkill/rfkill.hash
> new file mode 100644
> index 0000000..21be0a2
> --- /dev/null
> +++ b/package/rfkill/rfkill.hash
> @@ -0,0 +1,2 @@
> +# From: https://www.kernel.org/pub/software/network/rfkill/sha256sums.asc
> +sha256	e0ae3004215e39a6c5c36e0726558740728d16f67ebdb8bea621250f6091d86a  rfkill-0.5.tar.xz
> diff --git a/package/rfkill/rfkill.mk b/package/rfkill/rfkill.mk
> new file mode 100644
> index 0000000..b377c03
> --- /dev/null
> +++ b/package/rfkill/rfkill.mk
> @@ -0,0 +1,20 @@
> +################################################################################
> +#
> +# rfkill
> +#
> +################################################################################
> +
> +RFKILL_VERSION = 0.5
> +RFKILL_SOURCE = rfkill-$(RFKILL_VERSION).tar.xz
> +RFKILL_SITE = https://www.kernel.org/pub/software/network/rfkill
> +
> +define RFKILL_BUILD_CMDS
> +	$(TARGET_MAKE_ENV) $(TARGET_CONFIGURE_OPTS) $(MAKE) -C $(@D) \
> +		VERSION_SUFFIX="-br"

Why do you specify the VERSION_SUFFIX?

Don't do it, we do not need to "specialise" the generated binaries just
because they were built with Buldroot.

> +endef
> +
> +define RFKILL_INSTALL_TARGET_CMDS
> +	$(INSTALL) -D -m 755 $(@D)/rfkill $(TARGET_DIR)/usr/bin/rfkill

As I said in my previous review, you should use the install directive
provided by the package's Makefile, like so;

    $(TARGET_MAKE_ENV) $(TARGET_CONFIGURE_OPTS) $(MAKE) -C "$(@D) \
        DESTDIR=$(TARGET_DIR) install

Again, wait a little bit before you resend (like, wait until tomorow).

Thanks! :-)

Regards,
Yann E. MORIN.

> +endef
> +
> +$(eval $(generic-package))
> -- 
> 2.6.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

Viacheslav, All,

On 2015-10-27 17:13 +0100, Yann E. MORIN spake thusly:
> Since the previous review, we discussed on IRC, and we discovered
> more issues with your patch, see below...

Vicente had even more comments on my coments. :-)

(Hence the reason why waiting after a review is important.)

> On 2015-10-27 18:32 +0300, Viacheslav Volkov spake thusly:
[--SNIP--]
> > diff --git a/package/rfkill/rfkill.mk b/package/rfkill/rfkill.mk
> > new file mode 100644
> > index 0000000..b377c03
> > --- /dev/null
> > +++ b/package/rfkill/rfkill.mk
> > @@ -0,0 +1,20 @@
> > +################################################################################
> > +#
> > +# rfkill
> > +#
> > +################################################################################
> > +
> > +RFKILL_VERSION = 0.5
> > +RFKILL_SOURCE = rfkill-$(RFKILL_VERSION).tar.xz
> > +RFKILL_SITE = https://www.kernel.org/pub/software/network/rfkill
> > +
> > +define RFKILL_BUILD_CMDS
> > +	$(TARGET_MAKE_ENV) $(TARGET_CONFIGURE_OPTS) $(MAKE) -C $(@D) \
> > +		VERSION_SUFFIX="-br"
> 
> Why do you specify the VERSION_SUFFIX?
> 
> Don't do it, we do not need to "specialise" the generated binaries just
> because they were built with Buldroot.

OK, Vicente explained that without it, the version.sh script will fail
miserably. I understand now.

This should be explained as a comment above the build command:

    # As part of the build, the Maefile tries to construct a version
    # string. By default, it tries to run git commands to get it, and
    # checks that it gets the same version as the one stored in the
    # Makefile. Since it would (at worst) get Buildroot version (for
    # in-tree builds), or (at best) not version string at all (in case
    # Of out-of-tree build), the version.sh script would fail miserably.
    # To avoid this, we feed it a custom VERSION_SUFFIX, which must not
    # be empty, otherwise the version.sh script is run.

> > +endef
> > +
> > +define RFKILL_INSTALL_TARGET_CMDS
> > +	$(INSTALL) -D -m 755 $(@D)/rfkill $(TARGET_DIR)/usr/bin/rfkill
> 
> As I said in my previous review, you should use the install directive
> provided by the package's Makefile, like so;
> 
>     $(TARGET_MAKE_ENV) $(TARGET_CONFIGURE_OPTS) $(MAKE) -C "$(@D) \
>         DESTDIR=$(TARGET_DIR) install

The $(TARGET_CONFIGURE_OPTS) is not needed at install time. Vicente
believes t should be dropped; I believe it is better for consistency, so
the install command uses the exact same environment as the build
command.

So, please keep it. If a maintainer believes it is superfluous, he can
drop it the moment he applies the patch.

> Again, wait a little bit before you resend (like, wait until tomorow).

Still valid, by the way...

Thanks! :-)

Regards,
Yann E. MORIN.