[Buildroot] Fixing linux-pam build failures

[Buildroot] Fixing linux-pam build failures

[Dmitry Golubovsky]

Hi,

My initial assumption about the failures of linux-pam building in
autobuilder was not correct: the problem was in BerkeleyDB
configuration.

When BerkeleyDB is not built, the pam_userdb module is not built
either (via linux-pam configure). In this failed autobuilder
configuration Berkeley DB was selected. But dbm functionality was not
enabled in BerkeleyDB configure options. Thus the pam_userdb module
was attempted to build, but it failed due to the absence of dbm
functions in Berkeley DB, and this showed up in the build log.

I have submitted a series of patches which address the problem:

* if linux-pam is built, --enable-dbm is added to BerkeleyDB configure options
* BerkeleyDB is autoselected when linux-pam is selected, and
BerkeleyDB is now a dependency of linux-pam

I also included my two other patches (since linux-pam is being patched anyway):

* Provide dummy PAM config files to allow password-less logons via busybox login
* New configuration option for PAM support has been added to the
Busybox configuration menu.

Any suggestions are welcome.

Thanks.

-- 
Dmitry Golubovsky

Anywhere on the Web

[Buildroot] Fixing linux-pam build failures

[Stefan Fröberg]

Hi Dmitry!

Your work on this pam support looks good.

I have attached in this message my very quick, unpolished & very very
very dirty pam (version 1.1.5) hack
(it doesn't bother to even check for if pam is enabled in busybox and it
assumes that pam.d stuff is already in fs/skeleton/etc )

I can't even remember when I did this ugly hack.

Got it somehow working with the help of :
http://www.linuxfromscratch.org/blfs/view/svn/postlfs/linux-pam.html
http://www.linuxfromscratch.org/blfs/view/svn/postlfs/shadow.html
and
http://www.linuxfromscratch.org/blfs/view/svn/postlfs/cracklib.html

Those Linux From Scratch pages are really great starting templates for
any new buildroot package!
:)

Also, those three patches mentioned in this attached file are from
openwrt project witch also
uses buildroot  stuff (???).

https://dev.openwrt.org/browser/packages/libs/libpam

If you find something usefull for your work then by all means salvage
what you need from it.

I just want to get quickly rid of my dirty pam hack as soon as official
support lands on buildroot.

Stefan

P.S:

Your work for v86d ( http://patchwork.ozlabs.org/patch/157286/ )
is working wonderfully in my distro !
Thanks!


8.9.2012 7:32, Dmitry Golubovsky kirjoitti:
> Hi,
>
> My initial assumption about the failures of linux-pam building in
> autobuilder was not correct: the problem was in BerkeleyDB
> configuration.
>
> When BerkeleyDB is not built, the pam_userdb module is not built
> either (via linux-pam configure). In this failed autobuilder
> configuration Berkeley DB was selected. But dbm functionality was not
> enabled in BerkeleyDB configure options. Thus the pam_userdb module
> was attempted to build, but it failed due to the absence of dbm
> functions in Berkeley DB, and this showed up in the build log.
>
> I have submitted a series of patches which address the problem:
>
> * if linux-pam is built, --enable-dbm is added to BerkeleyDB configure options
> * BerkeleyDB is autoselected when linux-pam is selected, and
> BerkeleyDB is now a dependency of linux-pam
>
> I also included my two other patches (since linux-pam is being patched anyway):
>
> * Provide dummy PAM config files to allow password-less logons via busybox login
> * New configuration option for PAM support has been added to the
> Busybox configuration menu.
>
> Any suggestions are welcome.
>
> Thanks.
>

-------------- next part --------------

[Buildroot] Fixing linux-pam build failures

[Dmitry Golubovsky]

Stefan,

On Sat, Sep 8, 2012 at 7:28 AM, Stefan Fr?berg
<stefan.froberg@petroprogram.com> wrote:
> Hi Dmitry!
>
> Your work on this pam support looks good.
>
> I have attached in this message my very quick, unpolished & very very
> very dirty pam (version 1.1.5) hack
> (it doesn't bother to even check for if pam is enabled in busybox and it
> assumes that pam.d stuff is already in fs/skeleton/etc )

Thanks. However I'd rather let finalize the 1.1.4 integration (the
patches I just submitted will do) and then to upgrade to 1.1.5.

Another thing in your patch, and I was asked not to do so, is PAM
configs in the fs skeleton. I moved them to the Buildroot package
area, and honestly I would not provide any more configs with Buildroot
itself. For example, in my project PAM config for login does not
resemble the standard one at all, etc. these configs are just for
compatibility with PAM-less setup.

>
> If you find something usefull for your work then by all means salvage
> what you need from it.

Sure, once 1.1.4 is done.

> Your work for v86d ( http://patchwork.ozlabs.org/patch/157286/ )
> is working wonderfully in my distro !

Thanks for testing. The patchwork entry you mentioned also captures
the discussion we had regarding non-standard use of C flags: the way
Makefile is written in v86d does not capture them from the build
environment, and they have to be pushed harder down via redefinition
of $(CC) rather than via an environment variable.

I thnk eventually I'll try to reintroduce the v86d patch: may I take
your e-mail as a supporting vote? ;)

Thanks.

-- 
Dmitry Golubovsky

Anywhere on the Web

[Buildroot] Fixing linux-pam build failures

[Stefan Fröberg]

Hi Dmitry

8.9.2012 15:54, Dmitry Golubovsky kirjoitti:
>
>> Your work for v86d ( http://patchwork.ozlabs.org/patch/157286/ )
>> is working wonderfully in my distro !
> Thanks for testing. The patchwork entry you mentioned also captures
> the discussion we had regarding non-standard use of C flags: the way
> Makefile is written in v86d does not capture them from the build
> environment, and they have to be pushed harder down via redefinition
> of $(CC) rather than via an environment variable.
>
> I thnk eventually I'll try to reintroduce the v86d patch: may I take
> your e-mail as a supporting vote? ;)
>
> Thanks.
>
Of course.
I will be glad to :)

Stefan