value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

[Robert P. J. Day]

  just now noticed the variable PACKAGE_BEFORE_PN, whose entire
usage in poky (other than documentation) is:

meta/conf/bitbake.conf:PACKAGE_BEFORE_PN ?= ""
meta/conf/bitbake.conf:PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}"
meta/classes/lib_package.bbclass:PACKAGE_BEFORE_PN = "${PN}-bin"

  is there an actual need for that variable that places the new
package names *precisely* in that location in the PACKAGES list?  and
given the specificity of the initial PN-related packages:

  "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale"

is there any reason more recipes couldn't use PACKAGE_BEFORE_PN
instead of "PACKAGES =+" when defining new packages?  i *realize*
those two things don't represent the same thing but, in many cases,
would the end result be the same?

  is there any reason more code doesn't use "PACKAGE_BEFORE_PN"?

rday



-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================

Re: value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

[Saul Wold]

On 01/02/2013 09:04 AM, Robert P. J. Day wrote:
>
>    just now noticed the variable PACKAGE_BEFORE_PN, whose entire
> usage in poky (other than documentation) is:
>
> meta/conf/bitbake.conf:PACKAGE_BEFORE_PN ?= ""
> meta/conf/bitbake.conf:PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}"
> meta/classes/lib_package.bbclass:PACKAGE_BEFORE_PN = "${PN}-bin"
>
>    is there an actual need for that variable that places the new
> package names *precisely* in that location in the PACKAGES list?  and
> given the specificity of the initial PN-related packages:
>
Yes, the need in the case of lib_package it to ensure that the ${PN}-bin 
(bindir and sbindir) files get packages separately from the ${PN} files, 
which are typically the libraries in the case of recipes that deliver 
libraries and binaries.

>    "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale"
>
> is there any reason more recipes couldn't use PACKAGE_BEFORE_PN
> instead of "PACKAGES =+" when defining new packages?  i *realize*
> those two things don't represent the same thing but, in many cases,
> would the end result be the same?
>
>    is there any reason more code doesn't use "PACKAGE_BEFORE_PN"?
>
The =+ prepends it to the list and therefore is correctly greedy for the 
items they are trying to package. Since the PACKAGE_BEFORE_PN 
specifically inserts it, so for the most part there is no real need to 
use the P_B_P variable, and =+ is more consistent.

Sau!


> rday

Re: value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

[Robert P. J. Day]

On Wed, 2 Jan 2013, Saul Wold wrote:

> On 01/02/2013 09:04 AM, Robert P. J. Day wrote:
> >
> >    just now noticed the variable PACKAGE_BEFORE_PN, whose entire
> > usage in poky (other than documentation) is:
> >
> > meta/conf/bitbake.conf:PACKAGE_BEFORE_PN ?= ""
> > meta/conf/bitbake.conf:PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev
> > ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}"
> > meta/classes/lib_package.bbclass:PACKAGE_BEFORE_PN = "${PN}-bin"
> >
> >    is there an actual need for that variable that places the new
> > package names *precisely* in that location in the PACKAGES list?
> > and given the specificity of the initial PN-related packages:
> >
> Yes, the need in the case of lib_package it to ensure that the
> ${PN}-bin (bindir and sbindir) files get packages separately from
> the ${PN} files, which are typically the libraries in the case of
> recipes that deliver libraries and binaries.

  predictably, a few minutes after i posted, i realized what was going
on.  i was a bit confused since the variable name, PACKAGE_BEFORE_PN,
was so terrifically generic when its purpose is clearly to deal with
library packages for which (as you say) you'd like to break out the
generated binary artifacts under /usr/bin and /usr/sbin as a separate
output package, which makes sense.

> >    "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale"
> >
> > is there any reason more recipes couldn't use PACKAGE_BEFORE_PN
> > instead of "PACKAGES =+" when defining new packages?  i *realize*
> > those two things don't represent the same thing but, in many cases,
> > would the end result be the same?
> >
> >    is there any reason more code doesn't use "PACKAGE_BEFORE_PN"?
> >
> The =+ prepends it to the list and therefore is correctly greedy for
> the items they are trying to package. Since the PACKAGE_BEFORE_PN
> specifically inserts it, so for the most part there is no real need
> to use the P_B_P variable, and =+ is more consistent.

  the point i was trying to make above (probably badly) was that it
seemed unlikely there was much (effective) difference between, as you
point out, usage of P_B_P and using =+:

PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}"

unless those earlier packages in the above would greedily suck up any
of /usr/bin or /usr/sbin.

  so, to recap (to make sure i can explain it properly), OE's
bitbake.conf already defines this:

PACKAGE_BEFORE_PN ?= ""
PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}"
... snip ...
FILES_${PN}-bin = "${bindir}/* ${sbindir}/*"

which will have no effect unless, in your recipe, you inherit
lib_package, which pulls in:

PACKAGE_BEFORE_PN = "${PN}-bin"

  at which point, you can either accept the default definition of
FILES_${PN}-bin as sucking up *everything* left under /usr/bin and
/usr/sbin or you can set it explicitly to some other value, which
brings me to two final questions.

  is there any value in setting FILES_${PN}-bin without inheriting
lib_package?  see oe-core/meta/recipes-multimedia/pulseaudio, where
pulseaudio.inc contains the line:

FILES_${PN}-bin += "${sysconfdir}/default/volatiles/volatiles.04_pulse"

but there is no reference to lib_package anywhere in that recipe
directory.  so what's the point of that line above?

  on the other hand, you can always just set PACKAGE_BEFORE_PN
explicitly and skip all of lib_package.bbclass, like this in
meta-openembedded/meta-networking/recipes-protocols/quagga/quagga.inc:

# Split into a main package and separate per-protocol packages
PACKAGE_BEFORE_PN = "${PN}-ospfd ${PN}-ospf6d ${PN}-babeld ${PN}-bgpd
\
                     ${PN}-ripd ${PN}-ripngd ${PN}-isisd \
                     ${PN}-ospfclient ${PN}-watchquagga"

FILES_${PN}-ospfd       = "${sbindir}/ospfd ${libdir}/libospf.so.*"
FILES_${PN}-ospf6d      = "${sbindir}/ospf6d"
FILES_${PN}-babeld      = "${sbindir}/babeld"
FILES_${PN}-bgpd        = "${sbindir}/bgpd"
FILES_${PN}-ripd        = "${sbindir}/ripd"
FILES_${PN}-ripngd      = "${sbindir}/ripngd"
FILES_${PN}-isisd       = "${sbindir}/isisd"
FILES_${PN}-ospfclient  = "${sbindir}/ospfclient
${libdir}/libospfapiclient.so.*"
FILES_${PN}-watchquagga = "${sbindir}/watchquagga ${sysconfdir}/default/watchquagga \
                           ${sysconfdir}/init.d/watchquagga"


  is all of that about right?

rday

-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================

Re: value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

[Paul Eggleton]

On Thursday 03 January 2013 06:27:57 Robert P. J. Day wrote:
>   predictably, a few minutes after i posted, i realized what was going
> on.  i was a bit confused since the variable name, PACKAGE_BEFORE_PN,
> was so terrifically generic when its purpose is clearly to deal with
> library packages for which (as you say) you'd like to break out the
> generated binary artifacts under /usr/bin and /usr/sbin as a separate
> output package, which makes sense.

It does exactly what it says - puts package(s) before PN.

>   the point i was trying to make above (probably badly) was that it
> seemed unlikely there was much (effective) difference between, as you
> point out, usage of P_B_P and using =+:
> 
> PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale
> ${PACKAGE_BEFORE_PN} ${PN}"
> 
> unless those earlier packages in the above would greedily suck up any
> of /usr/bin or /usr/sbin.

That's correct. PACKAGE_BEFORE_PN just makes things a little easier.

>   is there any value in setting FILES_${PN}-bin without inheriting
> lib_package?

If ${PN}-bin has otherwise been added to PACKAGES, sure.

> see oe-core/meta/recipes-multimedia/pulseaudio, where
> pulseaudio.inc contains the line:
> 
> FILES_${PN}-bin += "${sysconfdir}/default/volatiles/volatiles.04_pulse"
> 
> but there is no reference to lib_package anywhere in that recipe
> directory.  so what's the point of that line above?

Looks like this is a mistake (or something left over from an earlier revision 
of the recipe) since ${PN}-bin isn't in PACKAGES for pulseaudio.

Cheers,
Paul
-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: value of "PACKAGE_BEFORE_PN" versus "PACKAGES =+"?

[Robert P. J. Day]

On Thu, 3 Jan 2013, Paul Eggleton wrote:

> On Thursday 03 January 2013 06:27:57 Robert P. J. Day wrote:
>
> > see oe-core/meta/recipes-multimedia/pulseaudio, where
> > pulseaudio.inc contains the line:
> >
> > FILES_${PN}-bin += "${sysconfdir}/default/volatiles/volatiles.04_pulse"
> >
> > but there is no reference to lib_package anywhere in that recipe
> > directory.  so what's the point of that line above?
>
> Looks like this is a mistake (or something left over from an earlier revision
> of the recipe) since ${PN}-bin isn't in PACKAGES for pulseaudio.

  i suspected something like that, just wanted to clarify.  i'll let
someone higher up the food chain fix that in whatever way is
appropriate.

rday

-- 

========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================