[PATCH] kernel.bbclass: add lzop dependency

[PATCH] kernel.bbclass: add lzop dependency

[Trevor Woerner]

If the initramfs image is type lzo, then a native lzop is needed.

Signed-off-by: Trevor Woerner <twoerner@gmail.com>
---
 meta/classes/kernel.bbclass | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/classes/kernel.bbclass b/meta/classes/kernel.bbclass
index fa76a14..db42744 100644
--- a/meta/classes/kernel.bbclass
+++ b/meta/classes/kernel.bbclass
@@ -1,7 +1,7 @@
 inherit linux-kernel-base kernel-module-split
 
 PROVIDES += "virtual/kernel"
-DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native"
+DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native lzop-native"
 
 S = "${STAGING_KERNEL_DIR}"
 B = "${WORKDIR}/build"
-- 
2.9.2

Re: [PATCH] kernel.bbclass: add lzop dependency

[Bruce Ashfield]

[-- Attachment #1: Type: text/plain, Size: 1426 bytes --]

On Mon, Aug 1, 2016 at 8:32 AM, Trevor Woerner <twoerner@gmail.com> wrote:

> If the initramfs image is type lzo, then a native lzop is needed.
>


Not a large dependency, but it does bring to mind the question if this
could be conditional on the type of image being built ? via  distro
feature, some other mechanism ?

Bruce


>
> Signed-off-by: Trevor Woerner <twoerner@gmail.com>
> ---
>  meta/classes/kernel.bbclass | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/meta/classes/kernel.bbclass b/meta/classes/kernel.bbclass
> index fa76a14..db42744 100644
> --- a/meta/classes/kernel.bbclass
> +++ b/meta/classes/kernel.bbclass
> @@ -1,7 +1,7 @@
>  inherit linux-kernel-base kernel-module-split
>
>  PROVIDES += "virtual/kernel"
> -DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc
> kmod-native depmodwrapper-cross bc-native"
> +DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc
> kmod-native depmodwrapper-cross bc-native lzop-native"
>
>  S = "${STAGING_KERNEL_DIR}"
>  B = "${WORKDIR}/build"
> --
> 2.9.2
>
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end"

[-- Attachment #2: Type: text/html, Size: 2389 bytes --]

Re: [PATCH] kernel.bbclass: add lzop dependency

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 404 bytes --]

On 1 August 2016 at 15:07, Bruce Ashfield <bruce.ashfield@gmail.com> wrote:

> Not a large dependency, but it does bring to mind the question if this
> could be conditional on the type of image being built ? via  distro
> feature, some other mechanism ?
>

Exactly this: image building adds dependencies automatically based on what
is actually being used, so can the kernel do this too?

Ross

[-- Attachment #2: Type: text/html, Size: 770 bytes --]

Re: [PATCH] kernel.bbclass: add lzop dependency

[Trevor Woerner]

On Mon 2016-08-01 @ 03:08:36 PM, Burton, Ross wrote:
> On 1 August 2016 at 15:07, Bruce Ashfield <bruce.ashfield@gmail.com> wrote:
> 
> > Not a large dependency, but it does bring to mind the question if this
> > could be conditional on the type of image being built ? via  distro
> > feature, some other mechanism ?
> >
> 
> Exactly this: image building adds dependencies automatically based on what
> is actually being used, so can the kernel do this too?

I'm uncertain regarding the status of this patch:

1) will it be applied? (doesn't sound like it)

2) do you want me to work on it some more? (i'll look for other examples of
conditional DEPENDS, i'm not aware of any off the top of my head)

3) should i be looking to push this into a BSP layer? (since BSPs usually
choose their kernel)

Re: [PATCH] kernel.bbclass: add lzop dependency

[Trevor Woerner]

On Mon 2016-08-01 @ 03:08:36 PM, Burton, Ross wrote:
> On 1 August 2016 at 15:07, Bruce Ashfield <bruce.ashfield@gmail.com> wrote:
> 
> > Not a large dependency, but it does bring to mind the question if this
> > could be conditional on the type of image being built ? via  distro
> > feature, some other mechanism ?
> >
> 
> Exactly this: image building adds dependencies automatically based on what
> is actually being used, so can the kernel do this too?

The way the build knows which image was built is to build the kernel then look
in the output directory to see what extension was given to the artifact. iow,
it only discovers the dependency partway through the build; it doesn't know
the dependency when the build starts.

Either bitbake will have to put together the kernel's config at the start of
the build, then parse through it to try to determine which kernel will be
built, or we need some sort of mechanism that allows us to add dependencies
partway through a build, or we can add (yet another) build variable and hope
people are able to keep it in sync.

...or we could just add lzop-native as a static dependency in the off-chance that
it's needed.

Re: [PATCH] kernel.bbclass: add lzop dependency

[Bruce Ashfield]

[-- Attachment #1: Type: text/plain, Size: 1894 bytes --]

On Wed, Aug 3, 2016 at 8:53 AM, Trevor Woerner <twoerner@gmail.com> wrote:

> On Mon 2016-08-01 @ 03:08:36 PM, Burton, Ross wrote:
> > On 1 August 2016 at 15:07, Bruce Ashfield <bruce.ashfield@gmail.com>
> wrote:
> >
> > > Not a large dependency, but it does bring to mind the question if this
> > > could be conditional on the type of image being built ? via  distro
> > > feature, some other mechanism ?
> > >
> >
> > Exactly this: image building adds dependencies automatically based on
> what
> > is actually being used, so can the kernel do this too?
>
> The way the build knows which image was built is to build the kernel then
> look
> in the output directory to see what extension was given to the artifact.
> iow,
> it only discovers the dependency partway through the build; it doesn't know
> the dependency when the build starts.
>
> Either bitbake will have to put together the kernel's config at the start
> of
> the build, then parse through it to try to determine which kernel will be
> built, or we need some sort of mechanism that allows us to add dependencies
> partway through a build, or we can add (yet another) build variable and
> hope
> people are able to keep it in sync.
>
>
I'm on vacation for the next two weeks, so sorry for the slow reply.

Wouldn't something like PACKAGECONFIG work here ? Something automatic
is going to be complex (like you describe), but something like this would
be an
easy way for a BSP (or other) layer to bbappend the kernel and add this as
a dependency.

... but then again, I suppose they could currently just bbappend the kernel
recipe
and add it directly to the DEPENDS.

Bruce


>
> ...or we could just add lzop-native as a static dependency in the
> off-chance that
> it's needed.
>



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end"

[-- Attachment #2: Type: text/html, Size: 2724 bytes --]

Re: [PATCH] kernel.bbclass: add lzop dependency

[Trevor Woerner]

On Wed 2016-08-03 @ 09:39:59 PM, Bruce Ashfield wrote:
> On Wed, Aug 3, 2016 at 8:53 AM, Trevor Woerner <twoerner@gmail.com> wrote:
> 
> > On Mon 2016-08-01 @ 03:08:36 PM, Burton, Ross wrote:
> > > On 1 August 2016 at 15:07, Bruce Ashfield <bruce.ashfield@gmail.com>
> > wrote:
> > >
> > > > Not a large dependency, but it does bring to mind the question if this
> > > > could be conditional on the type of image being built ? via  distro
> > > > feature, some other mechanism ?
> > > >
> > >
> > > Exactly this: image building adds dependencies automatically based on
> > what
> > > is actually being used, so can the kernel do this too?
> >
> > The way the build knows which image was built is to build the kernel then
> > look
> > in the output directory to see what extension was given to the artifact.
> > iow,
> > it only discovers the dependency partway through the build; it doesn't know
> > the dependency when the build starts.
> >
> > Either bitbake will have to put together the kernel's config at the start
> > of
> > the build, then parse through it to try to determine which kernel will be
> > built, or we need some sort of mechanism that allows us to add dependencies
> > partway through a build, or we can add (yet another) build variable and
> > hope
> > people are able to keep it in sync.
> >
> >
> I'm on vacation for the next two weeks, so sorry for the slow reply.

np enjoy! "'tis the season" :-)

> 
> Wouldn't something like PACKAGECONFIG work here ? Something automatic
> is going to be complex (like you describe), but something like this would
> be an
> easy way for a BSP (or other) layer to bbappend the kernel and add this as
> a dependency.
> 
> ... but then again, I suppose they could currently just bbappend the kernel
> recipe
> and add it directly to the DEPENDS.

I've proposed a change to the kernel.bbclass to add the lzop-native DEPENDS to
the existing

	DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native"

because it's in this file that the lzop program is being called.

It seems as though your suggestions are discussing changes to a recipe (?)
which is fine, I'll propose a patch to the BSP-in-question's kernel recipe.

> 
> >
> > ...or we could just add lzop-native as a static dependency in the
> > off-chance that
> > it's needed.
> >
> 
> 
> 
> -- 
> "Thou shalt not follow the NULL pointer, for chaos and madness await thee
> at its end"

Re: [PATCH] kernel.bbclass: add lzop dependency

[Mike Looijmans]

I've been adding "lzop-native" to a lot of kernel recipes because you also 
need it when generating an LZO compressed kernel (I like my boot time to be 
under 1 second).

It's a small dependency which is hard to automatically figure out, since there 
are a number of ways to trigger it (we just found two). I would welcome this 
patch very much, and I don't think many will suffer...


On 01-08-16 14:32, Trevor Woerner wrote:
> If the initramfs image is type lzo, then a native lzop is needed.
>
> Signed-off-by: Trevor Woerner <twoerner@gmail.com>
> ---
>   meta/classes/kernel.bbclass | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/meta/classes/kernel.bbclass b/meta/classes/kernel.bbclass
> index fa76a14..db42744 100644
> --- a/meta/classes/kernel.bbclass
> +++ b/meta/classes/kernel.bbclass
> @@ -1,7 +1,7 @@
>   inherit linux-kernel-base kernel-module-split
>
>   PROVIDES += "virtual/kernel"
> -DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native"
> +DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native lzop-native"
>
>   S = "${STAGING_KERNEL_DIR}"
>   B = "${WORKDIR}/build"
>

Re: [PATCH] kernel.bbclass: add lzop dependency

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1777 bytes --]


> On Aug 9, 2016, at 5:08 AM, Mike Looijmans <mike.looijmans@topic.nl> wrote:
> 
> I've been adding "lzop-native" to a lot of kernel recipes because you also need it when generating an LZO compressed kernel (I like my boot time to be under 1 second).
> 
> It's a small dependency which is hard to automatically figure out, since there are a number of ways to trigger it (we just found two). I would welcome this patch very much, and I don't think many will suffer…

From linux.inc days there is a python snippet which can do this conditionally e.g. see

https://github.com/kraj/meta-raspberrypi/blob/master/recipes-kernel/linux/linux-rpi.inc#L121

> 
> 
> On 01-08-16 14:32, Trevor Woerner wrote:
>> If the initramfs image is type lzo, then a native lzop is needed.
>> 
>> Signed-off-by: Trevor Woerner <twoerner@gmail.com>
>> ---
>>  meta/classes/kernel.bbclass | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/meta/classes/kernel.bbclass b/meta/classes/kernel.bbclass
>> index fa76a14..db42744 100644
>> --- a/meta/classes/kernel.bbclass
>> +++ b/meta/classes/kernel.bbclass
>> @@ -1,7 +1,7 @@
>>  inherit linux-kernel-base kernel-module-split
>> 
>>  PROVIDES += "virtual/kernel"
>> -DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native"
>> +DEPENDS += "virtual/${TARGET_PREFIX}binutils virtual/${TARGET_PREFIX}gcc kmod-native depmodwrapper-cross bc-native lzop-native"
>> 
>>  S = "${STAGING_KERNEL_DIR}"
>>  B = "${WORKDIR}/build"
>> 
> 
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core


[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]

Re: [PATCH] kernel.bbclass: add lzop dependency

[Richard Purdie]

On Tue, 2016-08-09 at 12:27 -0700, Khem Raj wrote:
> > On Aug 9, 2016, at 5:08 AM, Mike Looijmans <mike.looijmans@topic.nl
> > > wrote:
> > 
> > I've been adding "lzop-native" to a lot of kernel recipes because
> > you also need it when generating an LZO compressed kernel (I like
> > my boot time to be under 1 second).
> > 
> > It's a small dependency which is hard to automatically figure out,
> > since there are a number of ways to trigger it (we just found two).
> > I would welcome this patch very much, and I don't think many will
> > suffer…
> 
> From linux.inc days there is a python snippet which can do this
> conditionally e.g. see
> 
> https://github.com/kraj/meta-raspberrypi/blob/master/recipes-kernel/l
> inux/linux-rpi.inc#L121

Which is fine if you have a prebuild defconfig file however in many
cases we don't have that :(

Cheers,

Richard

Re: [PATCH] kernel.bbclass: add lzop dependency

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1129 bytes --]

On Aug 10, 2016 1:59 AM, "Richard Purdie" <
richard.purdie@linuxfoundation.org> wrote:
>
> On Tue, 2016-08-09 at 12:27 -0700, Khem Raj wrote:
> > > On Aug 9, 2016, at 5:08 AM, Mike Looijmans <mike.looijmans@topic.nl
> > > > wrote:
> > >
> > > I've been adding "lzop-native" to a lot of kernel recipes because
> > > you also need it when generating an LZO compressed kernel (I like
> > > my boot time to be under 1 second).
> > >
> > > It's a small dependency which is hard to automatically figure out,
> > > since there are a number of ways to trigger it (we just found two).
> > > I would welcome this patch very much, and I don't think many will
> > > suffer…
> >
> > From linux.inc days there is a python snippet which can do this
> > conditionally e.g. see
> >
> > https://github.com/kraj/meta-raspberrypi/blob/master/recipes-kernel/l
> > inux/linux-rpi.inc#L121
>
> Which is fine if you have a prebuild defconfig file however in many
> cases we don't have that :(
>

Thats right perhaps dependency could be conputed after configure task and
added to say compile task

> Cheers,
>
> Richard

[-- Attachment #2: Type: text/html, Size: 1691 bytes --]

Re: [PATCH] kernel.bbclass: add lzop dependency

[Richard Purdie]

On Wed, 2016-08-10 at 09:07 -0700, Khem Raj wrote:
> On Aug 10, 2016 1:59 AM, "Richard Purdie" <
> richard.purdie@linuxfoundation.org> wrote:
> >
> > On Tue, 2016-08-09 at 12:27 -0700, Khem Raj wrote:
> > > > On Aug 9, 2016, at 5:08 AM, Mike Looijmans <
> mike.looijmans@topic.nl
> > > > > wrote:
> > > >
> > > > I've been adding "lzop-native" to a lot of kernel recipes
> because
> > > > you also need it when generating an LZO compressed kernel (I
> like
> > > > my boot time to be under 1 second).
> > > >
> > > > It's a small dependency which is hard to automatically figure
> out,
> > > > since there are a number of ways to trigger it (we just found
> two).
> > > > I would welcome this patch very much, and I don't think many
> will
> > > > suffer…
> > >
> > > From linux.inc days there is a python snippet which can do this
> > > conditionally e.g. see
> > >
> > > 
> https://github.com/kraj/meta-raspberrypi/blob/master/recipes-kernel/l
> > > inux/linux-rpi.inc#L121
> >
> > Which is fine if you have a prebuild defconfig file however in many
> > cases we don't have that :(
> >
> Thats right perhaps dependency could be conputed after configure task
> and added to say compile task

Bitbake computes the dependency trees in advance, not on the fly so its
sadly not possible to change the dependency graph during a running
build. I'm sure you could design such a system but its not what we have
today, nor does it likely warrant the extra complexity it could
involve.

Cheers,

Richard