[OE-core] [PATCH 1/1] image.bbclass: add ROOTFS_POSTPROCESS_COMMAND to rootfs[vardeps]

[Jacob Kroon]

On Wed, 9 Jul 2014, Richard Purdie wrote:

> On Wed, 2014-07-09 at 12:08 +0300, Roxana Ciobanu wrote:
>>     Changes to ROOTFS_POSTPROCESS_COMMAND require the rootfs
>>     for the image to be rebuilt.
>>
>>     [YOCTO #6502]
>>
>> Signed-off-by: Roxana Ciobanu <roxana.ciobanu at intel.com>
>> ---
>>  meta/classes/image.bbclass | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/meta/classes/image.bbclass b/meta/classes/image.bbclass
>> index a03b880..e088bdb 100644
>> --- a/meta/classes/image.bbclass
>> +++ b/meta/classes/image.bbclass
>> @@ -75,7 +75,7 @@ LDCONFIGDEPEND_libc-musl = ""
>>  do_rootfs[depends] += "makedevs-native:do_populate_sysroot virtual/fakeroot-native:do_populate_sysroot ${LDCONFIGDEPEND}"
>>  do_rootfs[depends] += "virtual/update-alternatives-native:do_populate_sysroot update-rc.d-native:do_populate_sysroot"
>>  do_rootfs[recrdeptask] += "do_packagedata"
>> -do_rootfs[vardeps] += "BAD_RECOMMENDATIONS NO_RECOMMENDATIONS"
>> +do_rootfs[vardeps] += "BAD_RECOMMENDATIONS NO_RECOMMENDATIONS ROOTFS_POSTPROCESS_COMMAND"
>>
>>  do_build[depends] += "virtual/kernel:do_deploy"
>
> Are there other variables we look at in the rootfs process we also need
> to list here?
>
> As background, the move to lib/oe has moved the variables outside the
> scope of bitbake's function parsing code so we need to handle that
> manually now...
>

I was hit by this recently, and talked to kergoth on IRC about it.
As he suggested, perhaps we should also do

ROOTFS_POSTPROCESS_COMMAND[func] = "1"

to inform bitbake that it is a shell script, so that it invokes its 
shell dependency parsing. I would guess most people add a function, 
and append this to ROOTFS_POSTPROCESS_COMMAND. Without marking it as a 
function, changes in the actual function body would not trigger rootfs 
rebuilds.

/Jacob