Skip to contentSkip to conversation

curious about why bitbake.conf setting of FILES_${PN}-bin

DISCUSSION Robert P. J. Day 3 participants 6 messages Jul 15, 2014

lore
Share & tools
lore
#0 lore raw

currently doing a writeup on file distribution among a recipe's
generated packages, and noticed the following. here's a snippet from OE's bitbake.conf:

PACKAGE_BEFORE_PN ?= ""
PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}" PACKAGES_DYNAMIC = "^${PN}-locale-.*"
FILES = ""

FILES_${PN} = "${bindir}/* ${sbindir}/* ${libexecdir}/* ${libdir}/lib*${SOLIBS} \
            ${sysconfdir} ${sharedstatedir} ${localstatedir} \
            ${base_bindir}/* ${base_sbindir}/* \
            ${base_libdir}/*${SOLIBS} \
            ${base_prefix}/lib/udev/rules.d ${prefix}/lib/udev/rules.d \
            ${datadir}/${BPN} ${libdir}/${BPN}/* \
            ${datadir}/pixmaps ${datadir}/applications \
            ${datadir}/idl ${datadir}/omf ${datadir}/sounds \
            ${libdir}/bonobo/servers"

first, to make sure i understand the above correctly, the setting of
FILES_${PN} defines the (default) entire possible set of generated files that will be used to populate the packages created by a single recipe, correct?

also, since packages are populated in order, left to right, we'll
see file definitions like:

FILES_${PN}-dbg = ...
FILES_${PN}-staticdev = ...
FILES_${PN}-dev = ...

where, once a file is placed in a package, even if that name occurs again in a later package, it will be skipped. (anyone remember which manual this is mentioned in?)

however, i also see this:

FILES_${PN}-bin = "${bindir}/* ${sbindir}/*"

and i thought, that's weird, that particular package isn't mentioned anywhere in bitbake.conf, why is it being defined if it isn't used? ah, then i see this in lib_package.bbclass:

PACKAGE_BEFORE_PN = "${PN}-bin"

which clearly defines a library being packaged, but also allowing binary executables to be broken out separately, which is fine, but it's confusing why the setting of FILES_${PN}-bin is done in bitbake.conf, when its only application is (currently) for library packaging.

wouldn't it make more sense to move that line so that
lib_package.bbclass contained:

FILES_${PN}-bin = "${bindir}/* ${sbindir}/*"
PACKAGE_BEFORE_PN = "${PN}-bin"

that would make lib_package.bbclass more self-contained, and stop bitbake.conf from setting a variable that most recipes don't care about. thoughts?

rday

p.s. this kind of goes back to the image vs core-image discussion, where one wonders why base classes are doing things that require inheriting classes to finish off for them. or something like that.

Signature · 7 lines
========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================
#1 lore raw
Quoted 22 lines · Robert P. J. Day

currently doing a writeup on file distribution among a recipe's
generated packages, and noticed the following. here's a snippet from OE's bitbake.conf:

PACKAGE_BEFORE_PN ?= ""
PACKAGES = "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}" PACKAGES_DYNAMIC = "^${PN}-locale-.*"
FILES = ""

FILES_${PN} = "${bindir}/* ${sbindir}/* ${libexecdir}/* ${libdir}/lib*${SOLIBS} \
             ${sysconfdir} ${sharedstatedir} ${localstatedir} \
             ${base_bindir}/* ${base_sbindir}/* \
             ${base_libdir}/*${SOLIBS} \
             ${base_prefix}/lib/udev/rules.d ${prefix}/lib/udev/rules.d \
             ${datadir}/${BPN} ${libdir}/${BPN}/* \
             ${datadir}/pixmaps ${datadir}/applications \
             ${datadir}/idl ${datadir}/omf ${datadir}/sounds \
             ${libdir}/bonobo/servers"
Quoted 5 lines

first, to make sure i understand the above correctly, the setting of
FILES_${PN} defines the (default) entire possible set of generated files that will be used to populate the packages created by a single recipe, correct?

This is correct

Quoted 11 lines

also, since packages are populated in order, left to right, we'll
see file definitions like:

FILES_${PN}-dbg = ...
FILES_${PN}-staticdev = ...
FILES_${PN}-dev = ...

where, once a file is placed in a package, even if that name occurs again in a later package, it will be skipped. (anyone remember which manual this is mentioned in?)

I just did a basic search and could not find any reference and I know that PACKAGES/FILES is "greedy" meaning once a file is consumed by a FILES entry it's not available again. Should probably be added to the PACKAGES and / or FILES.

Quoted 19 lines

however, i also see this:

FILES_${PN}-bin = "${bindir}/* ${sbindir}/*"

and i thought, that's weird, that particular package isn't mentioned anywhere in bitbake.conf, why is it being defined if it isn't used? ah, then i see this in lib_package.bbclass:

PACKAGE_BEFORE_PN = "${PN}-bin"

which clearly defines a library being packaged, but also allowing binary executables to be broken out separately, which is fine, but it's confusing why the setting of FILES_${PN}-bin is done in bitbake.conf, when its only application is (currently) for library packaging.

wouldn't it make more sense to move that line so that
lib_package.bbclass contained:

Quoted 7 lines

FILES_${PN}-bin = "${bindir}/* ${sbindir}/*"
PACKAGE_BEFORE_PN = "${PN}-bin"

that would make lib_package.bbclass more self-contained, and stop bitbake.conf from setting a variable that most recipes don't care about. thoughts?

Blame me for that patch! Looking back at the commit message from about 2 years ago I think this got mid-way through a change, with the plan to actually remove lib_package completely since all it contained was the setting of PACKAGE_BEFORE_PN, therefore having FILES${PN}-bin. Not sure why we did not end up completing that.

So another option is to remove the inherits and replace it with PACKAGE_BEFORE_PNs, Which would finish off the orignal plan!

Quoted 6 lines

rday

p.s. this kind of goes back to the image vs core-image discussion, where one wonders why base classes are doing things that require inheriting classes to finish off for them. or something like that.

#2 lore raw
Quoted 4 lines · Saul Wold

I just did a basic search and could not find any reference and I know that PACKAGES/FILES is "greedy" meaning once a file is consumed by a FILES entry it's not available again. Should probably be added to the PACKAGES and / or FILES.

i *know* i've read that before somewhere, just don't remember where.
the only reference i've seen recently is, coincidentally, lib_package.bbclass:

#
# ${PN}-bin is defined in bitbake.conf
#
# We need to allow the other packages to be greedy with what they

# want out of /usr/bin and /usr/sbin before ${PN}-bin gets greedy.
#
PACKAGE_BEFORE_PN = "${PN}-bin"

anyone know if it's somewhere in one of the yocto manuals?

rday

Signature · 7 lines
========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

Twitter:                                       http://twitter.com/rpjday
LinkedIn:                               http://ca.linkedin.com/in/rpjday
========================================================================
#3 lore raw
In reply to Saul Wold
Quoted 8 lines · Saul Wold

Blame me for that patch! Looking back at the commit message from about 2 years ago I think this got mid-way through a change, with the plan to actually remove lib_package completely since all it contained was the setting of PACKAGE_BEFORE_PN, therefore having FILES${PN}-bin. Not sure why we did not end up completing that.

So another option is to remove the inherits and replace it with PACKAGE_BEFORE_PNs, Which would finish off the orignal plan!

Wouldn't that change the packaging by putting all binaries into a PN-bin package?

Ross

#4 lore raw
Quoted 7 lines · Ross
Quoted 6 lines · Saul Wold

Blame me for that patch! Looking back at the commit message from about 2 years ago I think this got mid-way through a change, with the plan to actually remove lib_package completely since all it contained was the setting of PACKAGE_BEFORE_PN, therefore having FILES${PN}-bin. Not sure why we did not end up completing that.

Quoted 6 lines
Quoted 2 lines

So another option is to remove the inherits and replace it with PACKAGE_BEFORE_PNs, Which would finish off the orignal plan!

Wouldn't that change the packaging by putting all binaries into a PN-bin package?

No the change would be to rmove the inherit lib_packaging from the recipes that use it and replace that with setting PACKAGE_BEFORE_PN with ${PN}-bin. So it's just doing what the recipes do already, not a general change in bitbake.conf.

Sau!

Quoted 3 lines

Ross

#5 lore raw
Quoted 12 lines · Saul Wold
Quoted 11 lines · Ross
Quoted 9 lines · Saul Wold

Blame me for that patch! Looking back at the commit message from about 2 years ago I think this got mid-way through a change, with the plan to actually remove lib_package completely since all it contained was the setting of PACKAGE_BEFORE_PN, therefore having FILES${PN}-bin. Not sure why we did not end up completing that.

So another option is to remove the inherits and replace it with PACKAGE_BEFORE_PNs, Which would finish off the orignal plan!

Quoted 7 lines
Quoted 3 lines

Wouldn't that change the packaging by putting all binaries into a PN-bin package?

No the change would be to rmove the inherit lib_packaging from the recipes that use it and replace that with setting PACKAGE_BEFORE_PN with ${PN}-bin. So it's just doing what the recipes do already, not a general change in bitbake.conf.

just FYI, while PACKAGE_BEFORE_PN is a cool concept, it's not like
there's a lot of it being used in the poky layer -- all of two uses:

meta/classes/lib_package.bbclass:PACKAGE_BEFORE_PN = "${PN}-bin" meta/recipes-graphics/clutter/clutter-1.0.inc:PACKAGE_BEFORE_PN += "${PN}-examples"

was it really necessary to introduce a new variable just for that?

rday

Signature · 7 lines
========================================================================
Robert P. J. Day                                 Ottawa, Ontario, CANADA
                        http://crashcourse.ca

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