[Buildroot] [PATCH 1/2] package/systemd: allow empty SYSTEMD_DEFAULT_TARGET

[Buildroot] [PATCH 1/2] package/systemd: allow empty SYSTEMD_DEFAULT_TARGET

[Norbert Lange]

Once you allow systemd to be build while not using it as init
system, the variable will be mpty and the build fails.

Signed-off-by: Norbert Lange <nolange79@gmail.com>
---
 package/systemd/systemd.mk | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/package/systemd/systemd.mk b/package/systemd/systemd.mk
index b6fffc553f..1510437617 100644
--- a/package/systemd/systemd.mk
+++ b/package/systemd/systemd.mk
@@ -586,7 +586,7 @@ SYSTEMD_CONF_OPTS += -Dfallback-hostname=$(SYSTEMD_FALLBACK_HOSTNAME)
 endif
 
 define SYSTEMD_INSTALL_INIT_HOOK
-	ln -fs "$(call qstrip,$(BR2_PACKAGE_SYSTEMD_DEFAULT_TARGET))" \
+	ln -fs "$(or $(call qstrip,$(BR2_PACKAGE_SYSTEMD_DEFAULT_TARGET)),multi-user.target)" \
 		$(TARGET_DIR)/usr/lib/systemd/system/default.target
 endef
 
-- 
2.35.1

_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

[Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Norbert Lange]

Allow systemd to be built as normal package, this for example
allows building systemd-boot without any init system.

Signed-off-by: Norbert Lange <nolange79@gmail.com>
---
 package/systemd/Config.in | 1 -
 1 file changed, 1 deletion(-)

diff --git a/package/systemd/Config.in b/package/systemd/Config.in
index 9f1a778e59..e0f5dbce0d 100644
--- a/package/systemd/Config.in
+++ b/package/systemd/Config.in
@@ -16,7 +16,6 @@ config BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
 
 menuconfig BR2_PACKAGE_SYSTEMD
 	bool "systemd"
-	depends on BR2_INIT_SYSTEMD
 	depends on BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
 	depends on BR2_USE_MMU
 	depends on !BR2_STATIC_LIBS # kmod
-- 
2.35.1

_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 1/2] package/systemd: allow empty SYSTEMD_DEFAULT_TARGET

[Yann E. MORIN]

Norbert, All,

On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
> Once you allow systemd to be build while not using it as init
> system, the variable will be mpty and the build fails.

A better commit log would be:

    Currently, we forcefully require that a default target be specified.
    However, systemd does install a default target, but it's not
    possible to just use that target.

    Allow the user to blank-out the config setting to not override
    whatever systemd installed and keep that.

(I don't think we want to be able to build systemd when it is not
selected as an init system, so we should not refer to that in the commit
log; see my reply to your next patch...)

> Signed-off-by: Norbert Lange <nolange79@gmail.com>
> ---
>  package/systemd/systemd.mk | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/package/systemd/systemd.mk b/package/systemd/systemd.mk
> index b6fffc553f..1510437617 100644
> --- a/package/systemd/systemd.mk
> +++ b/package/systemd/systemd.mk
> @@ -586,7 +586,7 @@ SYSTEMD_CONF_OPTS += -Dfallback-hostname=$(SYSTEMD_FALLBACK_HOSTNAME)
>  endif
>  
>  define SYSTEMD_INSTALL_INIT_HOOK
> -	ln -fs "$(call qstrip,$(BR2_PACKAGE_SYSTEMD_DEFAULT_TARGET))" \
> +	ln -fs "$(or $(call qstrip,$(BR2_PACKAGE_SYSTEMD_DEFAULT_TARGET)),multi-user.target)" \
>  		$(TARGET_DIR)/usr/lib/systemd/system/default.target
>  endef

Instead, the whole hook should be made conditional to whether the default
target is to be overriden:

    SYSTEMD_DEFAULT_TARGET = $(call qstrip,$(BR2_PACKAGE_SYSTEMD_DEFAULT_TARGET))
    ifneq ($(SYSTEMD_DEFAULT_TARGET),)
    define SYSTEMD_INSTALL_FORCE_TARGET
        ln -fs $(SYSTEMD_DEFAULT_TARGET) \
            $(TARGET_DIR)/usr/lib/systemd/system/default.target
    endef
    SYSTEMD_POST_INSTALL_TARGET_HOOKS += SYSTEMD_INSTALL_FORCE_TARGET
    endif

... and drop the assignment a few lines later, of course.

Regards,
Yann E. MORIN.

> -- 
> 2.35.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Yann E. MORIN]

Norbert, All,

On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
> Allow systemd to be built as normal package, this for example
> allows building systemd-boot without any init system.

How has the situation with upstream evolved since last time the topic
was raised here:
    https://lore.kernel.org/buildroot/cover.1545814334.git.yann.morin.1998@free.fr/#t
and especially the sub-thread:
    https://lore.kernel.org/buildroot/d38239f86f99924e87dcb6e045b2de3ac0cb435f.1545814334.git.yann.morin.1998@free.fr/

As far as I know (and understand), upstream has decided to tie the build
of systemd-boot (formerly the standalone gummiboot) to the build of the
rest of the systemd suite; they did not accept to make it possible to
only build systemd-boot to use it outside the systemd ecosystem.

From Buildroot's perspective, allowing to build systemd without it being
an init system, and without the possibility to just build systemd-boot,
is asking for troubles down the line.

The triviality of your patch does not make it possible to only just
build systemd-boot; instead, it still builds the whole systemd suite,
which hints that upstream did not make it possible to do separate
builds.

So, unless things have evolved substantially, I am not very fond of
alowing this, and stand by the decision that was made back then (i.e.
abide by upstream decision and tie systemd-boot to the systemd
ecosystem).

Regards,
Yann E. MORIN.

> Signed-off-by: Norbert Lange <nolange79@gmail.com>
> ---
>  package/systemd/Config.in | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/package/systemd/Config.in b/package/systemd/Config.in
> index 9f1a778e59..e0f5dbce0d 100644
> --- a/package/systemd/Config.in
> +++ b/package/systemd/Config.in
> @@ -16,7 +16,6 @@ config BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>  
>  menuconfig BR2_PACKAGE_SYSTEMD
>  	bool "systemd"
> -	depends on BR2_INIT_SYSTEMD
>  	depends on BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>  	depends on BR2_USE_MMU
>  	depends on !BR2_STATIC_LIBS # kmod
> -- 
> 2.35.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Norbert Lange]

Am Di., 19. Juli 2022 um 17:45 Uhr schrieb Yann E. MORIN
<yann.morin.1998@free.fr>:
>
> Norbert, All,
>
> On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
> > Allow systemd to be built as normal package, this for example
> > allows building systemd-boot without any init system.
>
> How has the situation with upstream evolved since last time the topic
> was raised here:
>     https://lore.kernel.org/buildroot/cover.1545814334.git.yann.morin.1998@free.fr/#t
> and especially the sub-thread:
>     https://lore.kernel.org/buildroot/d38239f86f99924e87dcb6e045b2de3ac0cb435f.1545814334.git.yann.morin.1998@free.fr/
>
> As far as I know (and understand), upstream has decided to tie the build
> of systemd-boot (formerly the standalone gummiboot) to the build of the
> rest of the systemd suite; they did not accept to make it possible to
> only build systemd-boot to use it outside the systemd ecosystem.
>
> From Buildroot's perspective, allowing to build systemd without it being
> an init system, and without the possibility to just build systemd-boot,
> is asking for troubles down the line.
>
> The triviality of your patch does not make it possible to only just
> build systemd-boot; instead, it still builds the whole systemd suite,
> which hints that upstream did not make it possible to do separate
> builds.

This was an example, and I do just that, even you built more than just
systemd-boot, picking "no init" speeds things up a bit.

> So, unless things have evolved substantially, I am not very fond of
> alowing this, and stand by the decision that was made back then (i.e.
> abide by upstream decision and tie systemd-boot to the systemd
> ecosystem).

I want to build systemd without using it as init system. Is there any
reason to artificially prohibit that?
It's just a regular package for me.

Norbert
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Arnout Vandecappelle]

On 21/07/2022 16:59, Norbert Lange wrote:
> Am Di., 19. Juli 2022 um 17:45 Uhr schrieb Yann E. MORIN
> <yann.morin.1998@free.fr>:
>>
>> Norbert, All,
>>
>> On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
>>> Allow systemd to be built as normal package, this for example
>>> allows building systemd-boot without any init system.
>>
>> How has the situation with upstream evolved since last time the topic
>> was raised here:
>>      https://lore.kernel.org/buildroot/cover.1545814334.git.yann.morin.1998@free.fr/#t
>> and especially the sub-thread:
>>      https://lore.kernel.org/buildroot/d38239f86f99924e87dcb6e045b2de3ac0cb435f.1545814334.git.yann.morin.1998@free.fr/
>>
>> As far as I know (and understand), upstream has decided to tie the build
>> of systemd-boot (formerly the standalone gummiboot) to the build of the
>> rest of the systemd suite; they did not accept to make it possible to
>> only build systemd-boot to use it outside the systemd ecosystem.
>>
>>  From Buildroot's perspective, allowing to build systemd without it being
>> an init system, and without the possibility to just build systemd-boot,
>> is asking for troubles down the line.
>>
>> The triviality of your patch does not make it possible to only just
>> build systemd-boot; instead, it still builds the whole systemd suite,
>> which hints that upstream did not make it possible to do separate
>> builds.
> 
> This was an example, and I do just that, even you built more than just
> systemd-boot, picking "no init" speeds things up a bit.
> 
>> So, unless things have evolved substantially, I am not very fond of
>> alowing this, and stand by the decision that was made back then (i.e.
>> abide by upstream decision and tie systemd-boot to the systemd
>> ecosystem).
> 
> I want to build systemd without using it as init system. Is there any
> reason to artificially prohibit that?
> It's just a regular package for me.

  AFAIU, systemd will unconditionally install /sbin/init, which will conflict 
with the other init system you have defined. So to complete this patch you 
should make the prompt depend on BR2_INIT_NONE.

  Same reasoning for udev (possible conflict with eudev), I'm not sure how to 
fix that one.

  There may be other conflicts as well, but the autobuilders will report those 
(once we re-enable the conflict check...).


  Regards,
  Arnout

> 
> Norbert
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Arnout Vandecappelle]

On 19/07/2022 17:45, Yann E. MORIN wrote:
> Norbert, All,
> 
> On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
>> Allow systemd to be built as normal package, this for example
>> allows building systemd-boot without any init system.
> 
> How has the situation with upstream evolved since last time the topic
> was raised here:
>      https://lore.kernel.org/buildroot/cover.1545814334.git.yann.morin.1998@free.fr/#t
> and especially the sub-thread:
>      https://lore.kernel.org/buildroot/d38239f86f99924e87dcb6e045b2de3ac0cb435f.1545814334.git.yann.morin.1998@free.fr/
> 
> As far as I know (and understand), upstream has decided to tie the build
> of systemd-boot (formerly the standalone gummiboot) to the build of the
> rest of the systemd suite; they did not accept to make it possible to
> only build systemd-boot to use it outside the systemd ecosystem.
> 
>  From Buildroot's perspective, allowing to build systemd without it being
> an init system, and without the possibility to just build systemd-boot,
> is asking for troubles down the line.

  However, it does make sense if you're not actually building a full system. 
That's not a use case that is really well supported by buildroot, but we 
shouldn't go out of our way to actively block that use case.

  I've been thinking for a while to make an explicit option for this, which 
would basically override/disable all the system/Config.in stuff. There's just 
one thing though: in such a use case, ou probably do still want to install the 
appropriate init scripts, so setting BR2_INIT_NONE is not exactly the correct 
either...

  Regards,
  Arnout

> 
> The triviality of your patch does not make it possible to only just
> build systemd-boot; instead, it still builds the whole systemd suite,
> which hints that upstream did not make it possible to do separate
> builds.
> 
> So, unless things have evolved substantially, I am not very fond of
> alowing this, and stand by the decision that was made back then (i.e.
> abide by upstream decision and tie systemd-boot to the systemd
> ecosystem).
> 
> Regards,
> Yann E. MORIN.
> 
>> Signed-off-by: Norbert Lange <nolange79@gmail.com>
>> ---
>>   package/systemd/Config.in | 1 -
>>   1 file changed, 1 deletion(-)
>>
>> diff --git a/package/systemd/Config.in b/package/systemd/Config.in
>> index 9f1a778e59..e0f5dbce0d 100644
>> --- a/package/systemd/Config.in
>> +++ b/package/systemd/Config.in
>> @@ -16,7 +16,6 @@ config BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>>   
>>   menuconfig BR2_PACKAGE_SYSTEMD
>>   	bool "systemd"
>> -	depends on BR2_INIT_SYSTEMD
>>   	depends on BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>>   	depends on BR2_USE_MMU
>>   	depends on !BR2_STATIC_LIBS # kmod
>> -- 
>> 2.35.1
>>
>> _______________________________________________
>> buildroot mailing list
>> buildroot@buildroot.org
>> https://lists.buildroot.org/mailman/listinfo/buildroot
> 
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/systemd: allow package to be used without init system

[Arnout Vandecappelle]

On 22/07/2022 23:29, Arnout Vandecappelle wrote:
>
>
> On 19/07/2022 17:45, Yann E. MORIN wrote:
>> Norbert, All,
>>
>> On 2022-07-19 15:16 +0200, Norbert Lange spake thusly:
>>> Allow systemd to be built as normal package, this for example
>>> allows building systemd-boot without any init system.
>>
>> How has the situation with upstream evolved since last time the topic
>> was raised here:
>> https://lore.kernel.org/buildroot/cover.1545814334.git.yann.morin.1998@free.fr/#t
>> and especially the sub-thread:
>> https://lore.kernel.org/buildroot/d38239f86f99924e87dcb6e045b2de3ac0cb435f.1545814334.git.yann.morin.1998@free.fr/
>>
>> As far as I know (and understand), upstream has decided to tie the build
>> of systemd-boot (formerly the standalone gummiboot) to the build of the
>> rest of the systemd suite; they did not accept to make it possible to
>> only build systemd-boot to use it outside the systemd ecosystem.
>>
>>  From Buildroot's perspective, allowing to build systemd without it being
>> an init system, and without the possibility to just build systemd-boot,
>> is asking for troubles down the line.
>
>  However, it does make sense if you're not actually building a full system. 
> That's not a use case that is really well supported by buildroot, but we 
> shouldn't go out of our way to actively block that use case.

  ... on the other hand, you can just select systemd as the init system even in 
that case. In the end, its only effects are:

- using the systemd skeleton rather than the empty skeleton;

- installing the systemd unit files for packages.

  The skeleton is not much of an overhead and is probably what you want anyway; 
the systemd unit files you probably want as well, as I observed before.


  So simply use BR2_INIT_SYSTEMD even if you're not actually using any init.


  Regards,
  Arnout



>
>  I've been thinking for a while to make an explicit option for this, which 
> would basically override/disable all the system/Config.in stuff. There's just 
> one thing though: in such a use case, ou probably do still want to install the 
> appropriate init scripts, so setting BR2_INIT_NONE is not exactly the correct 
> either...
>
>  Regards,
>  Arnout
>
>>
>> The triviality of your patch does not make it possible to only just
>> build systemd-boot; instead, it still builds the whole systemd suite,
>> which hints that upstream did not make it possible to do separate
>> builds.
>>
>> So, unless things have evolved substantially, I am not very fond of
>> alowing this, and stand by the decision that was made back then (i.e.
>> abide by upstream decision and tie systemd-boot to the systemd
>> ecosystem).
>>
>> Regards,
>> Yann E. MORIN.
>>
>>> Signed-off-by: Norbert Lange <nolange79@gmail.com>
>>> ---
>>>   package/systemd/Config.in | 1 -
>>>   1 file changed, 1 deletion(-)
>>>
>>> diff --git a/package/systemd/Config.in b/package/systemd/Config.in
>>> index 9f1a778e59..e0f5dbce0d 100644
>>> --- a/package/systemd/Config.in
>>> +++ b/package/systemd/Config.in
>>> @@ -16,7 +16,6 @@ config BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>>>     menuconfig BR2_PACKAGE_SYSTEMD
>>>       bool "systemd"
>>> -    depends on BR2_INIT_SYSTEMD
>>>       depends on BR2_PACKAGE_SYSTEMD_ARCH_SUPPORTS
>>>       depends on BR2_USE_MMU
>>>       depends on !BR2_STATIC_LIBS # kmod
>>> -- 
>>> 2.35.1
>>>
>>> _______________________________________________
>>> buildroot mailing list
>>> buildroot@buildroot.org
>>> https://lists.buildroot.org/mailman/listinfo/buildroot
>>
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot