[Buildroot] [PATCH 1/1] kmod: fix build with pkgconf 1.5.3

[Buildroot] [PATCH 1/1] kmod: fix build with pkgconf 1.5.3

[Fabrice Fontaine]

Since the bump to 1.5.3, pkgconf prepends the sysroot to all absolute
paths found in the .pc file. This is correct when the paths refer to
something in STAGING_DIR (e.g. libdir, includedir), but not when it
refers to something used for the target.

kmod uses the completionsdir variable from bash-completions.pc to decide
where to install things. Since DESTDIR is prepended to the install
destination, this will end up in the wrong location.

Until a better solution is found in pkgconf, pass the appdefaultdir to
use explicitly instead of relying on bash-completions.pc.

Fixes:
 - http://autobuild.buildroot.org/results/f8a1f956333062027294e766ff0ddab5c35d5887

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
---
 package/kmod/kmod.mk | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/package/kmod/kmod.mk b/package/kmod/kmod.mk
index c165ea3020..ea5d291133 100644
--- a/package/kmod/kmod.mk
+++ b/package/kmod/kmod.mk
@@ -28,6 +28,10 @@ KMOD_CONF_OPTS = --disable-static --enable-shared
 KMOD_CONF_OPTS += --disable-manpages
 HOST_KMOD_CONF_OPTS = --disable-manpages
 
+ifeq ($(BR2_PACKAGE_BASH_COMPLETION),y)
+KMOD_CONF_OPTS += --with-bashcompletiondir=/usr/share/bash-completion/completions
+endif
+
 ifeq ($(BR2_PACKAGE_ZLIB),y)
 KMOD_DEPENDENCIES += zlib
 KMOD_CONF_OPTS += --with-zlib
-- 
2.14.1

[Buildroot] [PATCH 1/1] kmod: fix build with pkgconf 1.5.3

[Peter Korsgaard]

>>>>> "Fabrice" == Fabrice Fontaine <fontaine.fabrice@gmail.com> writes:

 > Since the bump to 1.5.3, pkgconf prepends the sysroot to all absolute
 > paths found in the .pc file. This is correct when the paths refer to
 > something in STAGING_DIR (e.g. libdir, includedir), but not when it
 > refers to something used for the target.

 > kmod uses the completionsdir variable from bash-completions.pc to decide
 > where to install things. Since DESTDIR is prepended to the install
 > destination, this will end up in the wrong location.

 > Until a better solution is found in pkgconf, pass the appdefaultdir to
 > use explicitly instead of relying on bash-completions.pc.

 > Fixes:
 >  - http://autobuild.buildroot.org/results/f8a1f956333062027294e766ff0ddab5c35d5887

 > Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
 > ---
 >  package/kmod/kmod.mk | 4 ++++
 >  1 file changed, 4 insertions(+)

 > diff --git a/package/kmod/kmod.mk b/package/kmod/kmod.mk
 > index c165ea3020..ea5d291133 100644
 > --- a/package/kmod/kmod.mk
 > +++ b/package/kmod/kmod.mk
 > @@ -28,6 +28,10 @@ KMOD_CONF_OPTS = --disable-static --enable-shared
 >  KMOD_CONF_OPTS += --disable-manpages
 >  HOST_KMOD_CONF_OPTS = --disable-manpages
 
 > +ifeq ($(BR2_PACKAGE_BASH_COMPLETION),y)
 > +KMOD_CONF_OPTS += --with-bashcompletiondir=/usr/share/bash-completion/completions
 > +endif

Hmm, how did this work before if there wasn't anything to ensure
bash-completion was built before kmod? Just the alphabetical order?

Anyway, committed - Thanks.

-- 
Bye, Peter Korsgaard

[Buildroot] [PATCH 1/1] kmod: fix build with pkgconf 1.5.3

[Fabrice Fontaine]

Dear Peter,

Le ven. 14 d?c. 2018 ? 21:29, Peter Korsgaard <peter@korsgaard.com> a ?crit :
>
> >>>>> "Fabrice" == Fabrice Fontaine <fontaine.fabrice@gmail.com> writes:
>
>  > Since the bump to 1.5.3, pkgconf prepends the sysroot to all absolute
>  > paths found in the .pc file. This is correct when the paths refer to
>  > something in STAGING_DIR (e.g. libdir, includedir), but not when it
>  > refers to something used for the target.
>
>  > kmod uses the completionsdir variable from bash-completions.pc to decide
>  > where to install things. Since DESTDIR is prepended to the install
>  > destination, this will end up in the wrong location.
>
>  > Until a better solution is found in pkgconf, pass the appdefaultdir to
>  > use explicitly instead of relying on bash-completions.pc.
>
>  > Fixes:
>  >  - http://autobuild.buildroot.org/results/f8a1f956333062027294e766ff0ddab5c35d5887
>
>  > Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
>  > ---
>  >  package/kmod/kmod.mk | 4 ++++
>  >  1 file changed, 4 insertions(+)
>
>  > diff --git a/package/kmod/kmod.mk b/package/kmod/kmod.mk
>  > index c165ea3020..ea5d291133 100644
>  > --- a/package/kmod/kmod.mk
>  > +++ b/package/kmod/kmod.mk
>  > @@ -28,6 +28,10 @@ KMOD_CONF_OPTS = --disable-static --enable-shared
>  >  KMOD_CONF_OPTS += --disable-manpages
>  >  HOST_KMOD_CONF_OPTS = --disable-manpages
>
>  > +ifeq ($(BR2_PACKAGE_BASH_COMPLETION),y)
>  > +KMOD_CONF_OPTS += --with-bashcompletiondir=/usr/share/bash-completion/completions
>  > +endif
>
> Hmm, how did this work before if there wasn't anything to ensure
> bash-completion was built before kmod? Just the alphabetical order?
You're right, personally, I think that we should add a bash-completion
dependency but bash-completion is not a "real" build time dependency,
it's only needed to retrieve completionsdir.
Because I was told to remove udev dependency from alsa-utils during
review of https://patchwork.ozlabs.org/patch/1013180, I applied the
same logic for kmod.
Howerver if you want, I can send a new patch to add bash-completion dependency.
>
> Anyway, committed - Thanks.
>
> --
> Bye, Peter Korsgaard
Best Regards,

Fabrice

[Buildroot] [PATCH 1/1] kmod: fix build with pkgconf 1.5.3

[Peter Korsgaard]

>>>>> "Fabrice" == Fabrice Fontaine <fontaine.fabrice@gmail.com> writes:

Hi,

 >> Hmm, how did this work before if there wasn't anything to ensure
 >> bash-completion was built before kmod? Just the alphabetical order?

 > You're right, personally, I think that we should add a bash-completion
 > dependency but bash-completion is not a "real" build time dependency,
 > it's only needed to retrieve completionsdir.
 > Because I was told to remove udev dependency from alsa-utils during
 > review of https://patchwork.ozlabs.org/patch/1013180, I applied the
 > same logic for kmod.
 > Howerver if you want, I can send a new patch to add bash-completion dependency.

OK. As we now explicitly pass --with-completiondir (and hence not rely
on bash-completion.pc) and the install logic creates the destination
directory if needed, I don't think we really _NEED_ it.

-- 
Bye, Peter Korsgaard