[Buildroot] Buildroot and linux-pam?

[Buildroot] Buildroot and linux-pam?

[D M]

With regards to Dmitry's linux-pam paste-bin: we added linux-pam to our project, where
we use a CodeSourcery toolchain. ?We did not have to manually massage the config output
to get things to work, nor did we have to feed in the -lintl or -lfl. ?However, we did have
to add LINUX_PAM_AUTORECONF set to YES.  Otherwise, the compile eventually fails. ?
I was wondering if the LINUX_PAM_DISABLE_INNETGR?fixup could be avoided by doing?the reconfigure ...

Danomi -
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20120523/d5c7d3d5/attachment.html>

[Buildroot] Buildroot and linux-pam?

[Shawn Ferris]

> What did I do wrong? do you have any updates to this?

Nevermind.. as soon as I sent this, I realized I was wrong to think I
didn't have those functions. I removed the config updates made in this
package and it compiles

Shawn

[Buildroot] Buildroot and linux-pam?

[Shawn Ferris]

On Wed, May 16, 2012 at 6:55 AM, Dmitry Golubovsky <golubovsky@gmail.com> wrote:
> Hi,
>
> Thomas Petazzoni wrote:
>
> There are few functions (ruserok_af, innetgr) that uClibc is likely
> missing (or they are excluded per current configuration), and while
> calls to these functions ought to be disabled via configure, per
> existing code they are not. This might be a reportable bug*, but as my
> goal was to find a way to compile it in principle, I used such brute
> force replacing them with dummy macros just to get through. Another
> hack was overwriting makefiles in docs since Buildroot's toolchain may
> be missing some doc generation facilities, and I see little value in
> them anyway at least for Buildroot.

Hi Dmitry -- I was attempting to get systemd user sessions going when
I realized that pam was needed. I tried to work around it, but I'm
told it's futile. So, I saw this message and thought I'd give it a
shot. I added this package and updated my busybox and systemd packages
to reflect the addition. Upon compile I get this from linux-pam:

In file included from pam_access.c:45:0:
/root/build-root/panda-042012/output/host/usr/arm-unknown-linux-gnueabi/sysroot/usr/include/netdb.h:425:12:
error: expected identifier or ?(? before numeric constant
/root/build-root/panda-042012/output/host/usr/arm-unknown-linux-gnueabi/sysroot/usr/include/netdb.h:517:12:
error: expected identifier or ?(? before ?-? token

which of course is the definition of innetgr and ruserok_af
respectively. If I look at config.h, I see the additions:

#undef HAVE_RUSEROK
#define HAVE_RUSEROK_AF
#define ruserok_af(a, b, c, d, e) (-1)
#undef innetgr
#define innetgr(a, b, c, d) 0

What did I do wrong? do you have any updates to this?

Any help would be greatly appreciated!

Shawn

[Buildroot] Buildroot and linux-pam?

[Dmitry Golubovsky]

Hi,

Thomas Petazzoni wrote:

> You're apparently modifying the config.h manually, which is generated
> by ./configure script. Isn't it possible by passing environment
> variables to the ./configure script, to ensure that the problematic
> test gets the appropriate value?

There are few functions (ruserok_af, innetgr) that uClibc is likely
missing (or they are excluded per current configuration), and while
calls to these functions ought to be disabled via configure, per
existing code they are not. This might be a reportable bug*, but as my
goal was to find a way to compile it in principle, I used such brute
force replacing them with dummy macros just to get through. Another
hack was overwriting makefiles in docs since Buildroot's toolchain may
be missing some doc generation facilities, and I see little value in
them anyway at least for Buildroot.

Thanks for your comment.

------------------------
* But even if reported properly, it may take some time for it to be
reflected in a realase tarball, so perhaps for current adoption such
manual mending of config.h can be admitted.

-- 
Dmitry Golubovsky

Anywhere on the Web

[Buildroot] Buildroot and linux-pam?

[Thomas Petazzoni]

Hello Dmitry,

Le Wed, 16 May 2012 06:32:14 -0400,
Dmitry Golubovsky <golubovsky@gmail.com> a ?crit :

> It looks like linux-pam was never included in Buildroot, and there are
> few packages* where use of pam was explicitly disabled.
> 
> I managed to compile it with Buildroot though [1] with a number of
> hacks and workarounds. I tested it (at least pam_exec) with Qingy as a
> login program.
> 
> If I provide a better quality patch, will linux-pam be accepted? If
> yes, will there be any objections to (conditionally) enabling pam at
> least in systemd?

There will at least be no objection on my side, and in general, people
are fine with adding new features to existing packages.

> I am interested in both for my buildroot-based project which needs an
> authrntication facility. If for some reason linux-pam cannot be
> accepted, I'll keep this stuff as part of my project outside the
> buildroot tree.
> 
> Please let me know your opinion, or any suggestions how to improve the
> Makefile (apart from obvious cleanups).

We would definitely welcome libpam as a package. Your hacks don't look
really big, and there are probably ways to make them look a bit nicer.
You're apparently modifying the config.h manually, which is generated
by ./configure script. Isn't it possible by passing environment
variables to the ./configure script, to ensure that the problematic
test gets the appropriate value?

If you give a little bit more details on which problems those hacks are
solving, I'm sure we can help you to come up with solutions.

Thanks,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Buildroot and linux-pam?

[Dmitry Golubovsky]

Hi,

It looks like linux-pam was never included in Buildroot, and there are
few packages* where use of pam was explicitly disabled.

I managed to compile it with Buildroot though [1] with a number of
hacks and workarounds. I tested it (at least pam_exec) with Qingy as a
login program.

If I provide a better quality patch, will linux-pam be accepted? If
yes, will there be any objections to (conditionally) enabling pam at
least in systemd?

I am interested in both for my buildroot-based project which needs an
authrntication facility. If for some reason linux-pam cannot be
accepted, I'll keep this stuff as part of my project outside the
buildroot tree.

Please let me know your opinion, or any suggestions how to improve the
Makefile (apart from obvious cleanups).

Thanks.

----------------------------
* $ egrep -lr '(without|disable)-pam' package/
package/netatalk/netatalk.mk
package/systemd/systemd.mk
package/sudo/sudo.mk

[1] http://pastebin.com/jUWfXCMY (this code has not been cleaned up)

-- 
Dmitry Golubovsky

Anywhere on the Web