Re: [Ltib] rootn and sticky.

[Stuart Hughes]

Hi Bill,

I can see what you're trying to do, however this is at odds with trying
to keep the LTIB system and .spec files as simple as possible.  So as
far as it goes, it's not likely to get integrated back into the mainline
of LTIB.

You may want to investigate CONFIG_DISTRO (defaults.lkc and distro.lkc
under config/userspace), which could be used to select an alternative
set of .spec files.  This may help, but I don't think every really got
used before.

Regards, Stuart

On 23/04/12 18:15, Bill Pringlemeir wrote:
> I am looking for a way to create 'development' and 'deployment' rpms.
> The post sanitize rootfs -> rootfs.tmp works well if you only want to
> create a disk image.  I wanted to create distributable rpms.  This would
> allow a root files system to be updated using newer rpms.
>
> I have some spec files altered with different 'sub package' rules.  The
> prototype is something like,
>
>    [after normal package header]
>    %package devel
>    Summary: Development file for XXXXX.
>    Group: System Environment/Libraries
>    Provides: %{name}
>    %description devel
>    %{summary}
>    Development version.
>
>    [...]
>
>    %Files
>    %defattr(-,root,root)
>    # pared list of files.
>
>    # Full install for 'devel' version.
>    %Files devel
>    %defattr(-,root,root)
>    %{pfx}/*
>
> This creates a 'normal' rpm and a 'devel' rpm in rpm/RPMS/cpu.  I wanted
> to install the normal rpm to 'rootf.tmp' and the 'devel' rpm to
> 'rootfs'.  'normal' is not normal for ltib; I mean a normal user (as
> opposed to a developer).
>
> Currently I kludged 'build_rpm()' to install a 'devel' package if
> available and install the 'normal' package to 'roots.tmp'.  I had to
> alter f_clean(), and remove_unselected_pkgs() to handle the 'devel'
> sub packages as the 'rootfs' rpmdb contains these packages.
>
> It looked like '-rootn' might be trying to do the same sort of thing.
>
> The 'normal' rpm has some advantages.  It lists exactly what should be
> present and all MD5 sums can match with rpm verify (on the targets
> disk).  The size of the 'normal' rpm may be much smaller.  I think the
> 'remove man pages', etc could be accomplished with rpm macros in the
> main '%Files' section and some environment variables; but I haven't
> tried this yet.
>
> Surprisingly few projects actually need a 'devel' package to export
> headers for other to build; at least for the file system I am building.
>
> Thanks,
> Bill Pringlemeir.
>
> On 22 Apr 2012, address@hidden wrote:
>
>> Hi Bill,
>>
>> Those options are principally to support different root filesystem
>> images for CPUs with multiple cores.  This was stuff I was working on
>> before I left Freescale.  It was working, but not heavily tested.
>>
>> What is it you're trying to do?  Do you just want to specify where the
>> rootfs gets created, or something else?
>>
>> Regards, Stuart
>>
>> On 19/04/12 23:38, Bill Pringlemeir wrote:
>>> I have roamed around the Internet to find some info on this.  It looks
>>> like only LtibMergesFromFSL-20081205 has any information on this. Can we
>>> specify different 'rootfs' directories or does this mean something
>>> different?
>>>
>>> It looks like options such as --rootn -R -i --sticky and --no-sticky are
>>> not documented anywhere.  Are these touch me and I'll burn options?  I
>>> have used them and they didn't erase my hard drive.
>>>
>>> Thanks for any pointers.  Google 'ltib sticky OR rootn' doesn't give
>>> much help.
>>>
>>> Regards,
>>> Bill Pringlemeir.
>>>