[oe] [PATCH 1/4] lvm2: rebase all recipes on a global lvm2.inc recipe

[Simon Busch]

Am 28.07.2010 20:07, schrieb Stefan Schmidt:
> Hello.
> 
> On Wed, 2010-07-28 at 19:20, Simon Busch wrote:
>> On 28.07.2010 15:44, Stefan Schmidt wrote:
>>>
>>> On Mon, 2010-07-26 at 21:58, Simon Busch wrote:
>>>> On 26.07.2010 09:46, Stefan Schmidt wrote:
>>>>> Hello.
>>>>>
>>>>> On Tue, 2010-07-20 at 20:44, Simon Busch wrote:
>>>>>> This rebases all specific versions of lvm2 on a global recipe lvm2.inc which defines the
>>>>>> common parameters for building lvm2. Staging is overwritten as we don't need any of the
>>>>>> executables or manpages the build of lvm2 produces for any related builds.
>>>>>>
>>>>>> Signed-off-by: Simon Busch <morphis at gravedo.de>
>>>>>
>>>>> Looks good. Two comments below.
>>>>
>>>> Thank you for commenting this patches!
>>>>
>>>>>
>>>>>> diff --git a/recipes/lvm2/lvm2.inc b/recipes/lvm2/lvm2.inc
>>>>>> new file mode 100644
>>>>>> index 0000000..a7e37b5
>>>>>> --- /dev/null
>>>>>> +++ b/recipes/lvm2/lvm2.inc
>>>>>> @@ -0,0 +1,20 @@
>>>>>> +SECTION = "utils"
>>>>>> +DESCRIPTION = "LVM2 is a set of utilities to manage logical volumes in Linux."
>>>>>> +LICENSE = "GPL"
>>>>>> +DEPENDS = "device-mapper"
>>>>>> +INC_PR = "r2"
>>>>>> +
>>>>>> +S = "${WORKDIR}/LVM2.${PV}"
>>>>>> +SRC_URI = "ftp://sources.redhat.com/pub/lvm2/old/LVM2.${PV}.tgz \
>>>>>> +           file://crosscompile_fix.patch"
>>>>>> +
>>>>>> +# Unset user/group to unbreak install.
>>>>>> +EXTRA_OECONF = "--with-user= --with-group= --disable-o_direct"
>>>>>> +EXTRA_OECONF_arm = "--with-user= --with-group= --disable-o_direct"
>>>>>
>>>>> I can see that you just merged this in from an already existing recipe, but do
>>>>> we know why we have an overrirde for ARM here which changes nothing? Looks bogus
>>>>> to me.
>>>>
>>>> Hm, you are right. I checked it with recompiling without the
>>>> EXTRA_OECONF_arm statement. It works fine. Is it ok if I supply an extra
>>>> patch removing this statement or should I rework this patch?
>>>
>>> Both is fine with me.
>>>
>>>>>> +inherit autotools
>>>>>> +
>>>>>> +# We don't need to stage anything (the executables are no needed at build time by any
>>>>>> +# other recipe)
>>>>>> +do_stage() {
>>>>>> +}
>>>>>
>>>>> While we don't need the executables the don't hurd either IMHO. Can we get rid
>>>>> of this?
>>>>>
>>>>
>>>> The problem was, that the executables and manpages where installed into
>>>> the staging dir and never removed, when the packages was clean ur
>>>> purged. So I added the empty do_stage block to avoid stage of anything
>>>> from this package as we don't need them anyway.
>>>
>>> Hm, this smells like a bug somewhere else. Did the staged binaries bring in
>>> other problems or could they get staged until we find out why they are not
>>> removed?
>>
>> Staging this binaries let the recompile of the same package fail as it
>> tries to install the binaries + man pages again.
>> Any hints where the bug could be? Within the staging logic?
> 
> Good question. Maybe only a  bug in the clean step that it ignores the files
> during cleaning. Anybody else an idea?
> 
> You could also add a do_install_append which removes the binaries again.

Hm, is that the best solution? I think it's better to stage nothing as
we now we don't really need anything to be staged. So is it ok leaving
this patch as it is and commit it?

regards,
morphis