[RFC] Ship dbus config and dbus service activation files

[RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

While playing to get connman 0.1 running I found that we don't ship two files in
the wpa-supplicant that come in handy.

/etc/dbus-1/system.d/dbus-wpa_supplicant.conf

That one is really needed. Without it a wpa-supplicant -u barks out about wrong
security.

/usr/share/dbus-1/system-services/dbus-wpa_supplicant.service

This one is a nice-to-have. It allows autostart the service with the service
activation infrastructure from dbus. If a bus name listed in such a file is
called and not available it gets started the way described in this file.

The following patch i nstalls the files correctly for me. Never used FILES_${PN}
before so I would like to get some review before.

diff --git a/packages/wpa-supplicant/wpa-supplicant-0.6.inc b/packages/wpa-supplicant/wpa-supplicant-0.6.inc
index 6198958..779bee7 100644
--- a/packages/wpa-supplicant/wpa-supplicant-0.6.inc
+++ b/packages/wpa-supplicant/wpa-supplicant-0.6.inc
@@ -59,7 +59,14 @@ do_install () {
 	install -d ${D}${sysconfdir}/wpa_supplicant
 	install -m 755 ${WORKDIR}/ifupdown.sh ${D}${sysconfdir}/wpa_supplicant/
 	install -m 755 ${WORKDIR}/functions.sh ${D}${sysconfdir}/wpa_supplicant
-	
+
+	install -d ${D}/${sysconfdir}/dbus-1/system.d
+	install -m 644 ${S}/dbus-wpa_supplicant.conf ${D}/${sysconfdir}/dbus-1/system.d
+	install -d ${D}/${datadir}/dbus-1/system-services
+	install -m 644 ${S}/dbus-wpa_supplicant.service ${D}/${datadir}/dbus-1/system-services
+
 	ln -s /etc/wpa_supplicant/ifupdown.sh ${D}${sysconfdir}/network/if-pre-up.d/wpasupplicant
 	ln -s /etc/wpa_supplicant/ifupdown.sh ${D}${sysconfdir}/network/if-post-down.d/wpasupplicant
 }
+
+FILES_${PN} += "${datadir}/dbus-1/system-services/*"
diff --git a/packages/wpa-supplicant/wpa-supplicant_0.6.0.bb b/packages/wpa-supplicant/wpa-supplicant_0.6.0.bb
index d20ccaa..11fc774 100644
--- a/packages/wpa-supplicant/wpa-supplicant_0.6.0.bb
+++ b/packages/wpa-supplicant/wpa-supplicant_0.6.0.bb
@@ -1,3 +1,3 @@
 require wpa-supplicant-0.6.inc
 
-PR = "r1"
+PR = "r2"
diff --git a/packages/wpa-supplicant/wpa-supplicant_0.6.3.bb b/packages/wpa-supplicant/wpa-supplicant_0.6.3.bb
index d20ccaa..11fc774 100644
--- a/packages/wpa-supplicant/wpa-supplicant_0.6.3.bb
+++ b/packages/wpa-supplicant/wpa-supplicant_0.6.3.bb
@@ -1,3 +1,3 @@
 require wpa-supplicant-0.6.inc
 
-PR = "r1"
+PR = "r2"

Anyone objections?

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Thomas Kunze]

Stefan Schmidt schrieb:
> Hello.
>
> While playing to get connman 0.1 running I found that we don't ship two files in
> the wpa-supplicant that come in handy.
>
> /etc/dbus-1/system.d/dbus-wpa_supplicant.conf
>
> That one is really needed. Without it a wpa-supplicant -u barks out about wrong
> security.
>
> /usr/share/dbus-1/system-services/dbus-wpa_supplicant.service
>
> This one is a nice-to-have. It allows autostart the service with the service
> activation infrastructure from dbus. If a bus name listed in such a file is
> called and not available it gets started the way described in this file.
>
> The following patch i nstalls the files correctly for me. Never used FILES_${PN}
> before so I would like to get some review before.
>   
Maybe its better to create a package wpasupplicant-dbus. IMO we should 
try to split packages if we can and  the dbus
functionallity is clearly separable from wpasupplicant core functions.

Regards,
Thomas

Re: [RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

On Thu, 2008-12-11 at 01:16, Thomas Kunze wrote:
> Stefan Schmidt schrieb:

>> The following patch i nstalls the files correctly for me. Never used FILES_${PN}
>> before so I would like to get some review before.
>>   
> Maybe its better to create a package wpasupplicant-dbus. IMO we should  
> try to split packages if we can and  the dbus
> functionallity is clearly separable from wpasupplicant core functions.

I wouldn't mind. Where would you like to draw the line for such a split? Just
the dbus specific files or a wpa-supplicant version without and one with dbus
enabled?

If it is only about these two file it may not be worth the effort. I'm open
about this though.

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Michael 'Mickey' Lauer]

Am Thursday 11 December 2008 01:16:23 schrieb Thomas Kunze:
> > The following patch i nstalls the files correctly for me. Never used
> > FILES_${PN} before so I would like to get some review before.

Looks good to me.

>
> Maybe its better to create a package wpasupplicant-dbus. IMO we should
> try to split packages if we can and  the dbus
> functionallity is clearly separable from wpasupplicant core functions.

Is it? How so? AFAIK there is no facility to have wpa_supplicant dlopen 
libdbus and only provide the dbus interface on demand.

We can compile it completely without dbus support in a seperate recipe, if 
that's what you want -- but given the increasing ubiquity of dbus, that 
doesn't make much sense.

-- 
:M:

Re: [RFC] Ship dbus config and dbus service activation files

[Mike (mwester)]

Michael 'Mickey' Lauer wrote:
> We can compile it completely without dbus support in a seperate recipe, if 
> that's what you want -- but given the increasing ubiquity of dbus, that 
> doesn't make much sense.

Which raises a point -- there are still devices out there that don't
want the X11 libraries (due to flash space limitations) -- perhaps it is
_dbus_ that needs splitting into two separate recipes, one with X and
one without?

I think this would mean that everything would change to depend on
"virtual/dbus", right?  -- and that would be a fairly disruptive change,
hence my comments here to see if anyone has strong protests or better ideas.

Mike (mwester)

Re: [RFC] Ship dbus config and dbus service activation files

[Michael 'Mickey' Lauer]

Am Thursday 11 December 2008 01:55:39 schrieb Mike (mwester):
> Michael 'Mickey' Lauer wrote:
> > We can compile it completely without dbus support in a seperate recipe,
> > if that's what you want -- but given the increasing ubiquity of dbus,
> > that doesn't make much sense.
>
> Which raises a point -- there are still devices out there that don't
> want the X11 libraries (due to flash space limitations) -- perhaps it is
> _dbus_ that needs splitting into two separate recipes, one with X and
> one without?

I'd rather write a patch that tries to dlopen libX11. dbus needs libX11 only 
for setting one property on the root window.

-- 
:M:

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 01:55, Mike (mwester) wrote:
> Michael 'Mickey' Lauer wrote:
>> We can compile it completely without dbus support in a seperate recipe, if
>> that's what you want -- but given the increasing ubiquity of dbus, that
>> doesn't make much sense.
>
> Which raises a point -- there are still devices out there that don't
> want the X11 libraries (due to flash space limitations)

http://www.angstrom-distribution.org/repo/?pkgname=dbus-1
http://www.angstrom-distribution.org/repo/?pkgname=libdbus-1-3

Now where do you see it the _packages_ depending on libx11-6? Is dbus 
going to be the new python where everybody starts screaming about 
requiring X at runtime, when it's only a buildtime dep to build an 
*optional* subpackage like python-tk? Please stop with the FUD.

regards,

Koen

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 02:07, Michael 'Mickey' Lauer wrote:
> Am Thursday 11 December 2008 01:55:39 schrieb Mike (mwester):
>> Michael 'Mickey' Lauer wrote:
>>> We can compile it completely without dbus support in a seperate recipe,
>>> if that's what you want -- but given the increasing ubiquity of dbus,
>>> that doesn't make much sense.
>> Which raises a point -- there are still devices out there that don't
>> want the X11 libraries (due to flash space limitations) -- perhaps it is
>> _dbus_ that needs splitting into two separate recipes, one with X and
>> one without?
>
> I'd rather write a patch that tries to dlopen libX11. dbus needs libX11 only
> for setting one property on the root window.

Well, as I said in another mail, dbus-1 and libdbus-1-3 don't require 
x11 *at all*, so no patch needs to be written. For people wanting the 
x11 integration they can install the 'dbus-x11' subpackage. Just like 
python-tk :)

regards,

Koen

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 01:16, Thomas Kunze wrote:
> Stefan Schmidt schrieb:
>> Hello.
>>
>> While playing to get connman 0.1 running I found that we don't ship
>> two files in
>> the wpa-supplicant that come in handy.
>>
>> /etc/dbus-1/system.d/dbus-wpa_supplicant.conf
>>
>> That one is really needed. Without it a wpa-supplicant -u barks out
>> about wrong
>> security.
>>
>> /usr/share/dbus-1/system-services/dbus-wpa_supplicant.service
>>
>> This one is a nice-to-have. It allows autostart the service with the
>> service
>> activation infrastructure from dbus. If a bus name listed in such a
>> file is
>> called and not available it gets started the way described in this file.
>>
>> The following patch i nstalls the files correctly for me. Never used
>> FILES_${PN}
>> before so I would like to get some review before.
> Maybe its better to create a package wpasupplicant-dbus. IMO we should
> try to split packages if we can and the dbus
> functionallity is clearly separable from wpasupplicant core functions.

If it gets split (I don't have a strong opinion on that), don't forget 
to add an RDEPEND_${PN}-dbus = "dbus-1" to make it drag in dbus.

regards,

Koen

>
> Regards,
> Thomas

Re: [RFC] Ship dbus config and dbus service activation files

[Andrea Adami]

About connman, last time I built it I noticed it depends on
'resolvconf' which depends on 'bash'.

Is it right?


TIA

Andrea

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 10:37, Andrea Adami wrote:
> About connman, last time I built it I noticed it depends on
> 'resolvconf' which depends on 'bash'.
>
> Is it right?

It's right in so far that resolvconf is full of bashims. It would make 
more sense to use openresolv (http://roy.marples.name/node/343) to avoid 
dragging in bash. More to the point, OE's bash doesn't use u-a *and* 
doesn't register itself with /etc/shells, so resolvconf is broken.

regards,

Koen


>
>
> TIA
>
> Andrea

Re: [RFC] Ship dbus config and dbus service activation files

[Mike (mwester)]

Koen Kooi rudely writes:
> On 11-12-08 01:55, Mike (mwester) wrote:
>
> http://www.angstrom-distribution.org/repo/?pkgname=dbus-1
> http://www.angstrom-distribution.org/repo/?pkgname=libdbus-1-3
> 
> Now where do you see it the _packages_ depending on libx11-6? Is dbus
> going to be the new python where everybody starts screaming about
> requiring X at runtime, when it's only a buildtime dep to build an
> *optional* subpackage like python-tk? Please stop with the FUD.

dbus-1 recommends dbus-x11, dbus-x11 depends libx11-6 (and other X libs).

So there you go, opkg install <anything-depending-on-dbus> blows up your
flash space by pulling in numerous X libs.  No FUD - very real concern.
 (But if it's not a concern for Angstrom, then fine -- please ignore
this issue; if I fix it, I'll make sure that it won't change for Angstrom.)

Regards
Mike (mwester)

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 13:31, Mike (mwester) wrote:
> Koen Kooi rudely writes:
>> On 11-12-08 01:55, Mike (mwester) wrote:
>>
>> http://www.angstrom-distribution.org/repo/?pkgname=dbus-1
>> http://www.angstrom-distribution.org/repo/?pkgname=libdbus-1-3
>>
>> Now where do you see it the _packages_ depending on libx11-6? Is dbus
>> going to be the new python where everybody starts screaming about
>> requiring X at runtime, when it's only a buildtime dep to build an
>> *optional* subpackage like python-tk? Please stop with the FUD.
>
> dbus-1 recommends dbus-x11, dbus-x11 depends libx11-6 (and other X libs).
>
> So there you go, opkg install<anything-depending-on-dbus>  blows up your
> flash space by pulling in numerous X libs.  No FUD - very real concern.
>   (But if it's not a concern for Angstrom, then fine -- please ignore
> this issue; if I fix it, I'll make sure that it won't change for Angstrom.)

That rrecommends shouldn't be there specifically for the reasons you 
mentioned. I don't notice it since I looked up the old package name 
'dbus-1' instead of the new one 'dbus', sorry about that.

regards,

Koen

>
> Regards
> Mike (mwester)

Re: [RFC] Ship dbus config and dbus service activation files

[Koen Kooi]

On 11-12-08 13:31, Mike (mwester) wrote:
> Koen Kooi rudely writes:
>> On 11-12-08 01:55, Mike (mwester) wrote:
>>
>> http://www.angstrom-distribution.org/repo/?pkgname=dbus-1
>> http://www.angstrom-distribution.org/repo/?pkgname=libdbus-1-3
>>
>> Now where do you see it the _packages_ depending on libx11-6? Is dbus
>> going to be the new python where everybody starts screaming about
>> requiring X at runtime, when it's only a buildtime dep to build an
>> *optional* subpackage like python-tk? Please stop with the FUD.
>
> dbus-1 recommends dbus-x11, dbus-x11 depends libx11-6 (and other X libs).
>
> So there you go, opkg install<anything-depending-on-dbus>  blows up your
> flash space by pulling in numerous X libs.  No FUD - very real concern.
>   (But if it's not a concern for Angstrom, then fine -- please ignore
> this issue; if I fix it, I'll make sure that it won't change for Angstrom.)

Angstrom actually does the right things:

packagees/angstrom/angstrom-x11-base-depends.bb:  dbus-x11 \

It will drag in dbus-x11 when people want an image with an xserver in it.

regards,

Koen

Re: [RFC] Ship dbus config and dbus service activation files

[Andrea Adami]

Guys..don't get so rude...

Back to the topic:

>Anyone objections?
>
>regards
>Stefan Schmidt

Stefan, I think you have enough input to improve connman. This could
become a very nice proggy!

Regards
Andrea

Re: [RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

On Thu, 2008-12-11 at 10:37, Andrea Adami wrote:
> About connman, last time I built it I noticed it depends on
> 'resolvconf' which depends on 'bash'.
> 
> Is it right?

I don't remember if it was resolvconf or if it was dhclient, but one of it
pulled in bash. This may indeed be worth trying out the alternative Koen pointed
out.

This is low prio on my list until I get wpa-supplicant and therfor connman find
my device at all and start doing something usefull with it.

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

On Thu, 2008-12-11 at 01:33, Stefan Schmidt wrote:
> 
> On Thu, 2008-12-11 at 01:16, Thomas Kunze wrote:
> > Stefan Schmidt schrieb:
> 
> >> The following patch i nstalls the files correctly for me. Never used FILES_${PN}
> >> before so I would like to get some review before.
> >>   
> > Maybe its better to create a package wpasupplicant-dbus. IMO we should  
> > try to split packages if we can and  the dbus
> > functionallity is clearly separable from wpasupplicant core functions.
> 
> I wouldn't mind. Where would you like to draw the line for such a split? Just
> the dbus specific files or a wpa-supplicant version without and one with dbus
> enabled?

As we have sorted out the dbus depends on libX11 now it would be interesting if
people still think that separate version for wpa-supplicant, one with dbus and
one without, are still something to aim for?

The two files my patch adds are for sure not the problem here. I think nobody
will scream if I commit them.

Personally I would not step to do sucha package split as I don't have a system
that does not run dbus. (Which makes me abviously totally biased anyway. :) )

I'm fine on the other hand to test such a patch on the aspect if it still works
the way for us as before.

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

On Thu, 2008-12-11 at 14:06, Andrea Adami wrote:
> Guys..don't get so rude...

Hehe, that's ok. Getting used to rude email discussion over the years due to
various FOSS projects. :)

> Stefan, I think you have enough input to improve connman. This could
> become a very nice proggy!

I hope so. Will see what else needs fixing over the next weeks will getting it
running on the phone.

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Graeme Gregory]

On Thu, 2008-12-11 at 14:22 +0100, Stefan Schmidt wrote:
> Hello.
> 
> On Thu, 2008-12-11 at 10:37, Andrea Adami wrote:
> > About connman, last time I built it I noticed it depends on
> > 'resolvconf' which depends on 'bash'.
> > 
> > Is it right?
> 
> I don't remember if it was resolvconf or if it was dhclient, but one of it
> pulled in bash. This may indeed be worth trying out the alternative Koen pointed
> out.
> 
dhclient certainly used to use a script full of bashisms called when the
IP address is discovered. I had fun patching that one for NetworkManager

Graeme

Re: [RFC] Ship dbus config and dbus service activation files

[Stefan Schmidt]

Hello.

On Thu, 2008-12-11 at 13:38, Graeme Gregory wrote:
> On Thu, 2008-12-11 at 14:22 +0100, Stefan Schmidt wrote:
> > 
> > On Thu, 2008-12-11 at 10:37, Andrea Adami wrote:
> > > About connman, last time I built it I noticed it depends on
> > > 'resolvconf' which depends on 'bash'.
> > > 
> > > Is it right?
> > 
> > I don't remember if it was resolvconf or if it was dhclient, but one of it
> > pulled in bash. This may indeed be worth trying out the alternative Koen pointed
> > out.
> > 
> dhclient certainly used to use a script full of bashisms called when the
> IP address is discovered. I had fun patching that one for NetworkManager

Interesting. Your fixes are in OE.dev already?

I also got a pointer that maybe udhcp be worth looking into. Would need me to
look into a plugin for it. Still this has to wait until I got the basic
functions up and running.

regards
Stefan Schmidt

Re: [RFC] Ship dbus config and dbus service activation files

[Graeme Gregory]

On Thu, 2008-12-11 at 14:55 +0100, Stefan Schmidt wrote:
> Hello.
> 
> On Thu, 2008-12-11 at 13:38, Graeme Gregory wrote:
> > On Thu, 2008-12-11 at 14:22 +0100, Stefan Schmidt wrote:
> > > 
> > > On Thu, 2008-12-11 at 10:37, Andrea Adami wrote:
> > > > About connman, last time I built it I noticed it depends on
> > > > 'resolvconf' which depends on 'bash'.
> > > > 
> > > > Is it right?
> > > 
> > > I don't remember if it was resolvconf or if it was dhclient, but one of it
> > > pulled in bash. This may indeed be worth trying out the alternative Koen pointed
> > > out.
> > > 
> > dhclient certainly used to use a script full of bashisms called when the
> > IP address is discovered. I had fun patching that one for NetworkManager
> 
> Interesting. Your fixes are in OE.dev already?
> 
> I also got a pointer that maybe udhcp be worth looking into. Would need me to
> look into a plugin for it. Still this has to wait until I got the basic
> functions up and running.
> 
Yes, I used to use NetworkManager on my zaurus to manage the wifi!

Graeme