Re: [PATCH 1/2] linux-module: create S if not existing

[Anton Mikanovich <amikan@ilbers.de>]

08.07.2022 17:31, Moessbauer, Felix wrote:
>> -----Original Message-----
>> From: Kiszka, Jan (T CED) <jan.kiszka@siemens.com>
>> Sent: Thursday, July 7, 2022 6:54 PM
>> To: Schild, Henning (T CED SES-DE) <henning.schild@siemens.com>;
>> Moessbauer, Felix (T CED SES-DE) <felix.moessbauer@siemens.com>
>> Cc: isar-users@googlegroups.com
>> Subject: Re: [PATCH 1/2] linux-module: create S if not existing
>>
>> On 07.07.22 18:00, Henning Schild wrote:
>>> Am Thu,  7 Jul 2022 17:46:33 +0200
>>> schrieb Felix Moessbauer <felix.moessbauer@siemens.com>:
>>>
>>>> We copy the debian folder into S, but at do_prepare_build time S
>>>> might not have been created yet.
>>>> This patch makes sure that the directory is created if it does not
>>>> exist.
>>>>
>>>> Signed-off-by: Felix Moessbauer <felix.moessbauer@siemens.com>
>>>> ---
>>>>   meta/recipes-kernel/linux-module/module.inc | 1 +
>>>>   1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/meta/recipes-kernel/linux-module/module.inc
>>>> b/meta/recipes-kernel/linux-module/module.inc index
>>>> 50acfe14..97eff294 100644 ---
>>>> a/meta/recipes-kernel/linux-module/module.inc +++
>>>> b/meta/recipes-kernel/linux-module/module.inc @@ -28,6 +28,7 @@
>>>> TEMPLATE_FILES = "debian/control.tmpl \ debian/rules.tmpl"
>>>>   TEMPLATE_VARS += "KERNEL_NAME KERNEL_TYPE KERNEL_IMAGE_PKG
>>>> KERNEL_HEADERS_PKG PN"
>>>> +do_prepare_build[dirs] += "${S}"
>>>>   do_prepare_build() {
>>>>       cp -r ${WORKDIR}/debian ${S}/
>>> A little weird but i know the module you write that for is weird on
>>> its own ... nvidia. But what we do here is put packaging metadata into
>>> the code, which should be there after unpack ... everything else is "weird"
>>> ... no code no packaging needed.
>>>
>>> But hey it does not hurt.
>> It is definitely weird and will lead to new error patterns when folks specify a
>> wrong S value - did you check that?
> Well, not really.
> The module.inc already copies the Debian folder to S, so at worst in S will only be the Debian folder.
> Prior to this patch, this lead to an error message that the Debian folder could not be copied.
>
>> If a broken S will not silently generate a nop-recipe now, we can take this, but I
>> would like to see more reasoning of the case where S is filled with code AFTER
>> do_prepare_build.
> Yes, it might lead to a nop recipe, but we need a way to manually extract and move-around the artifacts that are downloaded to WORKDIR.
> Prior to this patch S has only been created by the do_unpack task when unpacking things into S.
>
> Having non-filled S is not so uncommon.
> Esp. in case the sources are provided by a Debian-binary package like "nvidia-kernel-source" which is added as a build dependency.
>
> Felix
>
>> Jan
>>
>> --
>> Siemens AG, Technology
>> Competence Center Embedded Linux

Can you provide recipe example for that case?
I think those weird recipes should be fixed by placing
do_prepare_build[dirs] += "${S}"
in downsteam recipe itself, but not in Isar.
Having incorrect ${S} is probably much more common case during development,
so we can't just allow it to be masked.