| Message ID | 20240807022718.24838-2-jose.fernandez@linux.dev (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [v2] kbuild: control extra pacman packages with PACMAN_EXTRAPACKAGES | expand |
On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > Introduce a new variable, PACMAN_EXTRAPACKAGES, in the Makefile.package > to control the creation of additional packages by the pacman-pkg target. > > The headers and api-headers packages will be included by default if > PACMAN_EXTRAPACKAGES is not set. This changes the previous behavior > where api-headers was always included, and headers was conditionally > included if CONFIG_MODULES=y. Now, this decision is delegated to the > user. > > To disable extra packages, set PACMAN_EXTRAPACKAGES to an empty value: > > make pacman-pkg PACMAN_EXTRAPACKAGES= > > or > > make pacman-pkg PACMAN_EXTRAPACKAGES="" > > Signed-off-by: Jose Fernandez <jose.fernandez@linux.dev> > Reviewed-by: Peter Jung <ptr1337@cachyos.org> > --- > v1 -> v2: Build all extra packages by default. Remove unnecessary lines. I see only the main package built by default. > > In a previous patch, there was concern that adding a new debug package > would increase the package time. To address this concern and provide > more flexibility, this change has been added to allow users to decide > which extra packages to include before introducing an optional debug > package [1]. > > [1] https://lore.kernel.org/lkml/20240801192008.GA3923315@thelio-3990X/T/ > > scripts/Makefile.package | 2 ++ > scripts/package/PKGBUILD | 11 +++++++---- > 2 files changed, 9 insertions(+), 4 deletions(-) > > diff --git a/scripts/Makefile.package b/scripts/Makefile.package > index 4a80584ec771..ccdf8ba41f0b 100644 > --- a/scripts/Makefile.package > +++ b/scripts/Makefile.package > @@ -144,6 +144,8 @@ snap-pkg: > # pacman-pkg > # --------------------------------------------------------------------------- > > +PACMAN_EXTRAPACKAGES ?= headers api-headers Meaningless line. Since 'export' is missing, this default line is not propagated to PKGBUILD. Nathan also mentioned 'export' would be needed if you wanted to describe this here. https://lore.kernel.org/linux-kbuild/20240806025853.GB1570554@thelio-3990X/ > + > PHONY += pacman-pkg > pacman-pkg: > @ln -srf $(srctree)/scripts/package/PKGBUILD $(objtree)/PKGBUILD > diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD > index 663ce300dd06..8de869f9b1d4 100644 > --- a/scripts/package/PKGBUILD > +++ b/scripts/package/PKGBUILD > @@ -3,10 +3,13 @@ > # Contributor: Jan Alexander Steffens (heftig) <heftig@archlinux.org> > > pkgbase=${PACMAN_PKGBASE:-linux-upstream} > -pkgname=("${pkgbase}" "${pkgbase}-api-headers") > -if grep -q CONFIG_MODULES=y include/config/auto.conf; then > - pkgname+=("${pkgbase}-headers") > -fi > +pkgname=("${pkgbase}") > + > +_extrapackages=${PACMAN_EXTRAPACKAGES:-} Instead of adding inconsistent defaults in two places, I would write like this: _extrapackages=${PACMAN_EXTRAPACKAGES-headers api-headers} Lastly, I will never accept new error messages with CONFIG_MODULES=n. > +for pkg in $_extrapackages; do > + pkgname+=("${pkgbase}-${pkg}") > +done > + > pkgver="${KERNELRELEASE//-/_}" > # The PKGBUILD is evaluated multiple times. > # Running scripts/build-version from here would introduce inconsistencies. > -- > 2.46.0 > -- Best Regards Masahiro Yamada
On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > Introduce a new variable, PACMAN_EXTRAPACKAGES, in the Makefile.package > > to control the creation of additional packages by the pacman-pkg target. > > > > The headers and api-headers packages will be included by default if > > PACMAN_EXTRAPACKAGES is not set. This changes the previous behavior > > where api-headers was always included, and headers was conditionally > > included if CONFIG_MODULES=y. Now, this decision is delegated to the > > user. > > > > To disable extra packages, set PACMAN_EXTRAPACKAGES to an empty value: > > > > make pacman-pkg PACMAN_EXTRAPACKAGES= > > > > or > > > > make pacman-pkg PACMAN_EXTRAPACKAGES="" > > > > Signed-off-by: Jose Fernandez <jose.fernandez@linux.dev> > > Reviewed-by: Peter Jung <ptr1337@cachyos.org> > > --- > > v1 -> v2: Build all extra packages by default. Remove unnecessary lines. > > > I see only the main package built by default. Same. Do we even need PACMAN_EXTRAPACKAGES in the Makefile? IMO having it in the PKGBUILD would be enough. It can still be overriden from the commandline. > > > > In a previous patch, there was concern that adding a new debug package > > would increase the package time. To address this concern and provide > > more flexibility, this change has been added to allow users to decide > > which extra packages to include before introducing an optional debug > > package [1]. > > > > [1] https://lore.kernel.org/lkml/20240801192008.GA3923315@thelio-3990X/T/ > > > > scripts/Makefile.package | 2 ++ > > scripts/package/PKGBUILD | 11 +++++++---- > > 2 files changed, 9 insertions(+), 4 deletions(-) > > > > diff --git a/scripts/Makefile.package b/scripts/Makefile.package > > index 4a80584ec771..ccdf8ba41f0b 100644 > > --- a/scripts/Makefile.package > > +++ b/scripts/Makefile.package > > @@ -144,6 +144,8 @@ snap-pkg: > > # pacman-pkg > > # --------------------------------------------------------------------------- > > > > +PACMAN_EXTRAPACKAGES ?= headers api-headers > > Meaningless line. > > > Since 'export' is missing, > this default line is not propagated to PKGBUILD. > > > Nathan also mentioned 'export' would be needed if you wanted to > describe this here. > > https://lore.kernel.org/linux-kbuild/20240806025853.GB1570554@thelio-3990X/ Same as above. > > + > > PHONY += pacman-pkg > > pacman-pkg: > > @ln -srf $(srctree)/scripts/package/PKGBUILD $(objtree)/PKGBUILD > > diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD > > index 663ce300dd06..8de869f9b1d4 100644 > > --- a/scripts/package/PKGBUILD > > +++ b/scripts/package/PKGBUILD > > @@ -3,10 +3,13 @@ > > # Contributor: Jan Alexander Steffens (heftig) <heftig@archlinux.org> > > > > pkgbase=${PACMAN_PKGBASE:-linux-upstream} > > -pkgname=("${pkgbase}" "${pkgbase}-api-headers") > > -if grep -q CONFIG_MODULES=y include/config/auto.conf; then > > - pkgname+=("${pkgbase}-headers") > > -fi > > +pkgname=("${pkgbase}") > > + > > +_extrapackages=${PACMAN_EXTRAPACKAGES:-} > > > Instead of adding inconsistent defaults in two places, > I would write like this: > > _extrapackages=${PACMAN_EXTRAPACKAGES-headers api-headers} Agreed. > Lastly, I will never accept new error messages > with CONFIG_MODULES=n. Could you elaborate? For me this works fine with CONFIG_MODULES=n. (After having fixed the above issues so all subpackages are built) > > +for pkg in $_extrapackages; do > > + pkgname+=("${pkgbase}-${pkg}") > > +done > > + > > pkgver="${KERNELRELEASE//-/_}" > > # The PKGBUILD is evaluated multiple times. > > # Running scripts/build-version from here would introduce inconsistencies. > > -- > > 2.46.0
On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > > > Introduce a new variable, PACMAN_EXTRAPACKAGES, in the Makefile.package > > > to control the creation of additional packages by the pacman-pkg target. > > > > > > The headers and api-headers packages will be included by default if > > > PACMAN_EXTRAPACKAGES is not set. This changes the previous behavior > > > where api-headers was always included, and headers was conditionally > > > included if CONFIG_MODULES=y. Now, this decision is delegated to the > > > user. > > > > > > To disable extra packages, set PACMAN_EXTRAPACKAGES to an empty value: > > > > > > make pacman-pkg PACMAN_EXTRAPACKAGES= > > > > > > or > > > > > > make pacman-pkg PACMAN_EXTRAPACKAGES="" > > > > > > Signed-off-by: Jose Fernandez <jose.fernandez@linux.dev> > > > Reviewed-by: Peter Jung <ptr1337@cachyos.org> > > > --- > > > v1 -> v2: Build all extra packages by default. Remove unnecessary lines. > > > > > > I see only the main package built by default. > > Same. > > Do we even need PACMAN_EXTRAPACKAGES in the Makefile? > IMO having it in the PKGBUILD would be enough. > It can still be overriden from the commandline. > > > > > > > In a previous patch, there was concern that adding a new debug package > > > would increase the package time. To address this concern and provide > > > more flexibility, this change has been added to allow users to decide > > > which extra packages to include before introducing an optional debug > > > package [1]. > > > > > > [1] https://lore.kernel.org/lkml/20240801192008.GA3923315@thelio-3990X/T/ > > > > > > scripts/Makefile.package | 2 ++ > > > scripts/package/PKGBUILD | 11 +++++++---- > > > 2 files changed, 9 insertions(+), 4 deletions(-) > > > > > > diff --git a/scripts/Makefile.package b/scripts/Makefile.package > > > index 4a80584ec771..ccdf8ba41f0b 100644 > > > --- a/scripts/Makefile.package > > > +++ b/scripts/Makefile.package > > > @@ -144,6 +144,8 @@ snap-pkg: > > > # pacman-pkg > > > # --------------------------------------------------------------------------- > > > > > > +PACMAN_EXTRAPACKAGES ?= headers api-headers > > > > Meaningless line. > > > > > > Since 'export' is missing, > > this default line is not propagated to PKGBUILD. > > > > > > Nathan also mentioned 'export' would be needed if you wanted to > > describe this here. > > > > https://lore.kernel.org/linux-kbuild/20240806025853.GB1570554@thelio-3990X/ > > Same as above. > > > > + > > > PHONY += pacman-pkg > > > pacman-pkg: > > > @ln -srf $(srctree)/scripts/package/PKGBUILD $(objtree)/PKGBUILD > > > diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD > > > index 663ce300dd06..8de869f9b1d4 100644 > > > --- a/scripts/package/PKGBUILD > > > +++ b/scripts/package/PKGBUILD > > > @@ -3,10 +3,13 @@ > > > # Contributor: Jan Alexander Steffens (heftig) <heftig@archlinux.org> > > > > > > pkgbase=${PACMAN_PKGBASE:-linux-upstream} > > > -pkgname=("${pkgbase}" "${pkgbase}-api-headers") > > > -if grep -q CONFIG_MODULES=y include/config/auto.conf; then > > > - pkgname+=("${pkgbase}-headers") > > > -fi > > > +pkgname=("${pkgbase}") > > > + > > > +_extrapackages=${PACMAN_EXTRAPACKAGES:-} > > > > > > Instead of adding inconsistent defaults in two places, > > I would write like this: > > > > _extrapackages=${PACMAN_EXTRAPACKAGES-headers api-headers} > > Agreed. > > > Lastly, I will never accept new error messages > > with CONFIG_MODULES=n. > > Could you elaborate? > For me this works fine with CONFIG_MODULES=n. > (After having fixed the above issues so all subpackages are built) $ make allnoconfig pacman-pkg Check the linux-headers log closely.
On 2024-08-08 02:02:59+0000, Masahiro Yamada wrote: > On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: <snip> > > > Lastly, I will never accept new error messages > > > with CONFIG_MODULES=n. > > > > Could you elaborate? > > For me this works fine with CONFIG_MODULES=n. > > (After having fixed the above issues so all subpackages are built) > > $ make allnoconfig pacman-pkg > > Check the linux-headers log closely. I see now, previously I was not on kbuild/for-next and had an old Module.symvers sitting around, hiding the issue. ==> Starting package_linux-upstream-headers()... Installing build files... tar: Module.symvers: Cannot stat: No such file or directory tar: Exiting with failure status due to previous errors Installing System.map and config... Adding symlink... ==> Tidying install... (coming from scripts/package/install-extmod-build) linux-upstream-headers also contains .config and System.map which are useful without modules. So either we completely disable linux-upstream-headers or skip install-extmod-build when CONFIG_MODULES=n. And maybe move System.map and .config to some other package, which would then deviate from the original PKGBUILD. Neither option feels great, but it probably won't make a big difference. If you have a preference, let's go with that. Thomas
On 24/08/07 07:31PM, Thomas Weißschuh wrote: > On 2024-08-08 02:02:59+0000, Masahiro Yamada wrote: > > On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > <snip> > > > > > Lastly, I will never accept new error messages > > > > with CONFIG_MODULES=n. > > > > > > Could you elaborate? > > > For me this works fine with CONFIG_MODULES=n. > > > (After having fixed the above issues so all subpackages are built) > > > > $ make allnoconfig pacman-pkg > > > > Check the linux-headers log closely. > > I see now, previously I was not on kbuild/for-next and had an old > Module.symvers sitting around, hiding the issue. > > ==> Starting package_linux-upstream-headers()... > Installing build files... > tar: Module.symvers: Cannot stat: No such file or directory > tar: Exiting with failure status due to previous errors > Installing System.map and config... > Adding symlink... > ==> Tidying install... > > (coming from scripts/package/install-extmod-build) > > linux-upstream-headers also contains .config and System.map which are > useful without modules. > So either we completely disable linux-upstream-headers or skip > install-extmod-build when CONFIG_MODULES=n. > And maybe move System.map and .config to some other package, > which would then deviate from the original PKGBUILD. > > Neither option feels great, but it probably won't make a big difference. > If you have a preference, let's go with that. Thomas, Masahiro, Thanks for the feedback. It seems that System.map and .config are commonly included in -header Arch packages. To avoid deviating too much and to address the issue with install-extmod-build when CONFIG_MODULES=n, how about considering something like this: mkdir -p "${builddir}" # needed if install-extmod-build is not run if grep -q CONFIG_MODULES=y include/config/auto.conf; then echo "Installing build files..." "${srctree}/scripts/package/install-extmod-build" "${builddir}" fi echo "Installing System.map and config..." cp System.map "${builddir}/System.map" cp .config "${builddir}/.config" Thanks, Jose
On Sat, Aug 10, 2024 at 9:16 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > On 24/08/07 07:31PM, Thomas Weißschuh wrote: > > On 2024-08-08 02:02:59+0000, Masahiro Yamada wrote: > > > On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > > > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > <snip> > > > > > > > Lastly, I will never accept new error messages > > > > > with CONFIG_MODULES=n. > > > > > > > > Could you elaborate? > > > > For me this works fine with CONFIG_MODULES=n. > > > > (After having fixed the above issues so all subpackages are built) > > > > > > $ make allnoconfig pacman-pkg > > > > > > Check the linux-headers log closely. > > > > I see now, previously I was not on kbuild/for-next and had an old > > Module.symvers sitting around, hiding the issue. > > > > ==> Starting package_linux-upstream-headers()... > > Installing build files... > > tar: Module.symvers: Cannot stat: No such file or directory > > tar: Exiting with failure status due to previous errors > > Installing System.map and config... > > Adding symlink... > > ==> Tidying install... > > > > (coming from scripts/package/install-extmod-build) > > > > linux-upstream-headers also contains .config and System.map which are > > useful without modules. > > So either we completely disable linux-upstream-headers or skip > > install-extmod-build when CONFIG_MODULES=n. > > And maybe move System.map and .config to some other package, > > which would then deviate from the original PKGBUILD. > > > > Neither option feels great, but it probably won't make a big difference. > > If you have a preference, let's go with that. > > Thomas, Masahiro, > Thanks for the feedback. It seems that System.map and .config are commonly > included in -header Arch packages. To avoid deviating too much and to address > the issue with install-extmod-build when CONFIG_MODULES=n, how about considering > something like this: I am fine. > > mkdir -p "${builddir}" # needed if install-extmod-build is not run This comment might not be necessary if you move this code right before copying System.map. > if grep -q CONFIG_MODULES=y include/config/auto.conf; then > echo "Installing build files..." > "${srctree}/scripts/package/install-extmod-build" "${builddir}" > fi > > echo "Installing System.map and config..." > cp System.map "${builddir}/System.map" > cp .config "${builddir}/.config" > > Thanks, > Jose
On 2024-08-09 18:16:36+0000, Jose Fernandez wrote: > On 24/08/07 07:31PM, Thomas Weißschuh wrote: > > On 2024-08-08 02:02:59+0000, Masahiro Yamada wrote: > > > On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > > > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > <snip> > > > > > > > Lastly, I will never accept new error messages > > > > > with CONFIG_MODULES=n. > > > > > > > > Could you elaborate? > > > > For me this works fine with CONFIG_MODULES=n. > > > > (After having fixed the above issues so all subpackages are built) > > > > > > $ make allnoconfig pacman-pkg > > > > > > Check the linux-headers log closely. > > > > I see now, previously I was not on kbuild/for-next and had an old > > Module.symvers sitting around, hiding the issue. > > > > ==> Starting package_linux-upstream-headers()... > > Installing build files... > > tar: Module.symvers: Cannot stat: No such file or directory > > tar: Exiting with failure status due to previous errors > > Installing System.map and config... > > Adding symlink... > > ==> Tidying install... > > > > (coming from scripts/package/install-extmod-build) > > > > linux-upstream-headers also contains .config and System.map which are > > useful without modules. > > So either we completely disable linux-upstream-headers or skip > > install-extmod-build when CONFIG_MODULES=n. > > And maybe move System.map and .config to some other package, > > which would then deviate from the original PKGBUILD. > > > > Neither option feels great, but it probably won't make a big difference. > > If you have a preference, let's go with that. > > Thomas, Masahiro, > Thanks for the feedback. It seems that System.map and .config are commonly > included in -header Arch packages. To avoid deviating too much and to address > the issue with install-extmod-build when CONFIG_MODULES=n, how about considering > something like this: > > mkdir -p "${builddir}" # needed if install-extmod-build is not run > if grep -q CONFIG_MODULES=y include/config/auto.conf; then > echo "Installing build files..." > "${srctree}/scripts/package/install-extmod-build" "${builddir}" > fi > > echo "Installing System.map and config..." > cp System.map "${builddir}/System.map" > cp .config "${builddir}/.config" Sounds good to me. Thomas
On 24/08/10 04:41PM, Masahiro Yamada wrote: > On Sat, Aug 10, 2024 at 9:16 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > On 24/08/07 07:31PM, Thomas Weißschuh wrote: > > > On 2024-08-08 02:02:59+0000, Masahiro Yamada wrote: > > > > On Thu, Aug 8, 2024 at 1:41 AM Thomas Weißschuh <linux@weissschuh.net> wrote: > > > > > On 2024-08-07 22:37:47+0000, Masahiro Yamada wrote: > > > > > > On Wed, Aug 7, 2024 at 11:28 AM Jose Fernandez <jose.fernandez@linux.dev> wrote: > > > > > > <snip> > > > > > > > > > Lastly, I will never accept new error messages > > > > > > with CONFIG_MODULES=n. > > > > > > > > > > Could you elaborate? > > > > > For me this works fine with CONFIG_MODULES=n. > > > > > (After having fixed the above issues so all subpackages are built) > > > > > > > > $ make allnoconfig pacman-pkg > > > > > > > > Check the linux-headers log closely. > > > > > > I see now, previously I was not on kbuild/for-next and had an old > > > Module.symvers sitting around, hiding the issue. > > > > > > ==> Starting package_linux-upstream-headers()... > > > Installing build files... > > > tar: Module.symvers: Cannot stat: No such file or directory > > > tar: Exiting with failure status due to previous errors > > > Installing System.map and config... > > > Adding symlink... > > > ==> Tidying install... > > > > > > (coming from scripts/package/install-extmod-build) > > > > > > linux-upstream-headers also contains .config and System.map which are > > > useful without modules. > > > So either we completely disable linux-upstream-headers or skip > > > install-extmod-build when CONFIG_MODULES=n. > > > And maybe move System.map and .config to some other package, > > > which would then deviate from the original PKGBUILD. > > > > > > Neither option feels great, but it probably won't make a big difference. > > > If you have a preference, let's go with that. > > > > Thomas, Masahiro, > > Thanks for the feedback. It seems that System.map and .config are commonly > > included in -header Arch packages. To avoid deviating too much and to address > > the issue with install-extmod-build when CONFIG_MODULES=n, how about considering > > something like this: > > > I am fine. > > > > > > mkdir -p "${builddir}" # needed if install-extmod-build is not run > > This comment might not be necessary if you move this code > right before copying System.map. Ack. I will move the command to before copying System.map and drop the comment. > > > if grep -q CONFIG_MODULES=y include/config/auto.conf; then > > echo "Installing build files..." > > "${srctree}/scripts/package/install-extmod-build" "${builddir}" > > fi > > > > echo "Installing System.map and config..." > > cp System.map "${builddir}/System.map" > > cp .config "${builddir}/.config" > > > > Thanks, > > Jose > > > > -- > Best Regards > Masahiro Yamada
On 24/08/07 06:40PM, Thomas Weißschuh wrote: <snip> > > > + > > > PHONY += pacman-pkg > > > pacman-pkg: > > > @ln -srf $(srctree)/scripts/package/PKGBUILD $(objtree)/PKGBUILD > > > diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD > > > index 663ce300dd06..8de869f9b1d4 100644 > > > --- a/scripts/package/PKGBUILD > > > +++ b/scripts/package/PKGBUILD > > > @@ -3,10 +3,13 @@ > > > # Contributor: Jan Alexander Steffens (heftig) <heftig@archlinux.org> > > > > > > pkgbase=${PACMAN_PKGBASE:-linux-upstream} > > > -pkgname=("${pkgbase}" "${pkgbase}-api-headers") > > > -if grep -q CONFIG_MODULES=y include/config/auto.conf; then > > > - pkgname+=("${pkgbase}-headers") > > > -fi > > > +pkgname=("${pkgbase}") > > > + > > > +_extrapackages=${PACMAN_EXTRAPACKAGES:-} > > > > > > Instead of adding inconsistent defaults in two places, > > I would write like this: > > > > _extrapackages=${PACMAN_EXTRAPACKAGES-headers api-headers} > > Agreed. > Closing the loop on this topic. I removed all changes made to Makefile.package and set the default in PKGBUILD as suggested: _extrapackages=${PACMAN_EXTRAPACKAGES-headers api-headers} Running the pacman-pkg target without setting PACMAN_EXTRAPACKAGES will build all packages: make pacman-pkg ... ==> Creating package "linux-upstream"... ... ==> Creating package "linux-upstream-headers"... ... ==> Creating package "linux-upstream-api-headers"... Setting PACMAN_EXTRAPACKAGES to an empty value will build only the kernel package: make pacman-pkg PACMAN_EXTRAPACKAGES="" objtree="/home/jose/Code/linux/linux" \ BUILDDIR="/home/jose/Code/linux/linux/pacman" \ CARCH="i386" \ KBUILD_MAKEFLAGS="rR --no-print-directory -- PACMAN_EXTRAPACKAGES=" \ KBUILD_REVISION="46" \ makepkg ... ==> Creating package "linux-upstream"... -> Generating .PKGINFO file... -> Generating .BUILDINFO file... -> Generating .MTREE file... -> Compressing package... ==> Leaving fakeroot environment. ==> Finished making: linux-upstream 6.11.0_rc2+-46 (Sun 11 Aug 2024 01:13:45 PM MDT) Make exports command line arguments as env variables to sub-processes and this is how the PACMAN_EXTRAPACKAGES is passed to makepg without an explicit export with this change. V3 will include this change.
diff --git a/scripts/Makefile.package b/scripts/Makefile.package index 4a80584ec771..ccdf8ba41f0b 100644 --- a/scripts/Makefile.package +++ b/scripts/Makefile.package @@ -144,6 +144,8 @@ snap-pkg: # pacman-pkg # --------------------------------------------------------------------------- +PACMAN_EXTRAPACKAGES ?= headers api-headers + PHONY += pacman-pkg pacman-pkg: @ln -srf $(srctree)/scripts/package/PKGBUILD $(objtree)/PKGBUILD diff --git a/scripts/package/PKGBUILD b/scripts/package/PKGBUILD index 663ce300dd06..8de869f9b1d4 100644 --- a/scripts/package/PKGBUILD +++ b/scripts/package/PKGBUILD @@ -3,10 +3,13 @@ # Contributor: Jan Alexander Steffens (heftig) <heftig@archlinux.org> pkgbase=${PACMAN_PKGBASE:-linux-upstream} -pkgname=("${pkgbase}" "${pkgbase}-api-headers") -if grep -q CONFIG_MODULES=y include/config/auto.conf; then - pkgname+=("${pkgbase}-headers") -fi +pkgname=("${pkgbase}") + +_extrapackages=${PACMAN_EXTRAPACKAGES:-} +for pkg in $_extrapackages; do + pkgname+=("${pkgbase}-${pkg}") +done + pkgver="${KERNELRELEASE//-/_}" # The PKGBUILD is evaluated multiple times. # Running scripts/build-version from here would introduce inconsistencies.