| Message ID | 20250714134657.1620685-1-martin.jansa@gmail.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [meta-oe] fwupd: set vendor_ids_dir | expand |
Hi, On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > From: Martin Jansa <martin.jansa@gmail.com> > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > meson is searching the paths on host with: > > # look for usb.ids in both of the Debian and Fedora locations, > # and fall back to the system datadir in case we're building in a venv or prefix > vendor_ids_dir = get_option('vendor_ids_dir') > if vendor_ids_dir == '' > vendor_ids_dir = join_paths(datadir, 'misc') > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > vendor_ids_dir = join_paths(datadir, 'hwdata') Isn't this the same as ${datadir}/hwdata ? Cheers, -Mikko > endif > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > vendor_ids_dir = '/usr/share/hwdata' > endif > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > vendor_ids_dir = '/usr/share/misc' > endif > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > vendor_ids_dir = '/usr/local/var/homebrew/linked/usb.ids/share/misc' > endif > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > error('could not auto-detect -Dvendor_ids_dir=') > endif > endif > conf.set_quoted ('FWUPD_DATADIR_VENDOR_IDS', vendor_ids_dir) > > it shouldn't be looking at host's paths and and on hosts without > usb.ids in one of these locations it fails with: > > ../sources/fwupd-2.0.12/meson.build:323:4: ERROR: Problem encountered: could not auto-detect -Dvendor_ids_dir= > > Signed-off-by: Martin Jansa <martin.jansa@gmail.com> > --- > meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb b/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb > index e25fbe1f54..eb43c3d865 100644 > --- a/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb > +++ b/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb > @@ -37,6 +37,8 @@ GIDOCGEN_MESON_DISABLE_FLAG = 'disabled' > GIR_MESON_ENABLE_FLAG = 'enabled' > GIR_MESON_DISABLE_FLAG = 'disabled' > > +EXTRA_OEMESON = "-Dvendor_ids_dir=${datadir}/hwdata" > + > PACKAGECONFIG ??= "\ > ${@bb.utils.contains('DISTRO_FEATURES', 'systemd', 'systemd', '', d)} \ > ${@bb.utils.contains('PTEST_ENABLED', '1', 'tests', '', d)} \
On Mon, Jul 14, 2025 at 3:52 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > Hi, > > On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > > From: Martin Jansa <martin.jansa@gmail.com> > > > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > > meson is searching the paths on host with: > > > > # look for usb.ids in both of the Debian and Fedora locations, > > # and fall back to the system datadir in case we're building in a venv or prefix > > vendor_ids_dir = get_option('vendor_ids_dir') > > if vendor_ids_dir == '' > > vendor_ids_dir = join_paths(datadir, 'misc') > > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > > vendor_ids_dir = join_paths(datadir, 'hwdata') > > Isn't this the same as ${datadir}/hwdata ? It is the same value, but on my host, I don't have /usr/share/hwdata/usb.ids so meson doesn't detect any existing vendor_ids_dir value and fails. With this change it accepts passed vendor_ids_dir value without testing it's existence (neither on host nor in RSS). Cheers,
Hi, On Mon, Jul 14, 2025 at 04:00:03PM +0200, Martin Jansa wrote: > On Mon, Jul 14, 2025 at 3:52 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > > > From: Martin Jansa <martin.jansa@gmail.com> > > > > > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > > > meson is searching the paths on host with: > > > > > > # look for usb.ids in both of the Debian and Fedora locations, > > > # and fall back to the system datadir in case we're building in a venv or prefix > > > vendor_ids_dir = get_option('vendor_ids_dir') > > > if vendor_ids_dir == '' > > > vendor_ids_dir = join_paths(datadir, 'misc') > > > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > > > vendor_ids_dir = join_paths(datadir, 'hwdata') > > > > Isn't this the same as ${datadir}/hwdata ? > > It is the same value, but on my host, I don't have > /usr/share/hwdata/usb.ids so meson doesn't detect any existing > vendor_ids_dir value and fails. > > With this change it accepts passed vendor_ids_dir value without > testing it's existence (neither on host nor in RSS). Previous version of fwupd update patches did not include "hwdata" as build dependency so recipe sysroot did not have usb.ids file but now it should have. Then meson.bbclass is setting datadir so I'm wondering if there is a bug in meson.bbclass or in these fwupd side scripts that they only check the host paths on not the recipe sysroot ones? Is fwupd not using the correct variables? Should the fwupd meson config actually check if the file is in join_paths(sys_root, datadir, 'hwdata')? I'm fine with this change but just wondering if there is something more missing. Cheers, -Mikko
On Mon, Jul 14, 2025 at 4:31 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > Hi, > > On Mon, Jul 14, 2025 at 04:00:03PM +0200, Martin Jansa wrote: > > On Mon, Jul 14, 2025 at 3:52 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > > On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > > > > From: Martin Jansa <martin.jansa@gmail.com> > > > > > > > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > > > > meson is searching the paths on host with: > > > > > > > > # look for usb.ids in both of the Debian and Fedora locations, > > > > # and fall back to the system datadir in case we're building in a venv or prefix > > > > vendor_ids_dir = get_option('vendor_ids_dir') > > > > if vendor_ids_dir == '' > > > > vendor_ids_dir = join_paths(datadir, 'misc') > > > > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > > > > vendor_ids_dir = join_paths(datadir, 'hwdata') > > > > > > Isn't this the same as ${datadir}/hwdata ? > > > > It is the same value, but on my host, I don't have > > /usr/share/hwdata/usb.ids so meson doesn't detect any existing > > vendor_ids_dir value and fails. > > > > With this change it accepts passed vendor_ids_dir value without > > testing it's existence (neither on host nor in RSS). > > Previous version of fwupd update patches did not include "hwdata" > as build dependency so recipe sysroot did not have usb.ids file and previous version wasn't failing like this, it started to fail only after the upgrade as https://github.com/fwupd/fwupd/commit/61a863de8dd429be8c393ec267d63db78de71b76#diff-30d8f6be6320feeacf686be94f48c70869b52630e01ea625f0f15adc0d57c3e4 was introduced in version 2. The hwdata dependency added in v2 might not be needed, I guess it should be runtime dependency for libfwupdplugin not build time. > but now it should have. Then meson.bbclass is setting datadir > so I'm wondering if there is a bug in meson.bbclass or in these fwupd > side scripts that they only check the host paths on not the > recipe sysroot ones? Is fwupd not using the correct variables? Yes it checks the paths on host not in RSS as said in commit message. > Should the fwupd meson config actually check if the file is in > join_paths(sys_root, datadir, 'hwdata')? Something like this might work, but not sure where you expect sys_root to be defined. I see it added as a property by meson.bbclass, but doing just this causes: ERROR: Unknown variable "sys_root".
On Mon, Jul 14, 2025 at 5:12 PM Martin Jansa <martin.jansa@gmail.com> wrote: > > On Mon, Jul 14, 2025 at 4:31 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > > > Hi, > > > > On Mon, Jul 14, 2025 at 04:00:03PM +0200, Martin Jansa wrote: > > > On Mon, Jul 14, 2025 at 3:52 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > > > On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > > > > > From: Martin Jansa <martin.jansa@gmail.com> > > > > > > > > > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > > > > > meson is searching the paths on host with: > > > > > > > > > > # look for usb.ids in both of the Debian and Fedora locations, > > > > > # and fall back to the system datadir in case we're building in a venv or prefix > > > > > vendor_ids_dir = get_option('vendor_ids_dir') > > > > > if vendor_ids_dir == '' > > > > > vendor_ids_dir = join_paths(datadir, 'misc') > > > > > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > > > > > vendor_ids_dir = join_paths(datadir, 'hwdata') > > > > > > > > Isn't this the same as ${datadir}/hwdata ? > > > > > > It is the same value, but on my host, I don't have > > > /usr/share/hwdata/usb.ids so meson doesn't detect any existing > > > vendor_ids_dir value and fails. > > > > > > With this change it accepts passed vendor_ids_dir value without > > > testing it's existence (neither on host nor in RSS). > > > > Previous version of fwupd update patches did not include "hwdata" > > as build dependency so recipe sysroot did not have usb.ids file > > and previous version wasn't failing like this, it started to fail only > after the upgrade as > https://github.com/fwupd/fwupd/commit/61a863de8dd429be8c393ec267d63db78de71b76#diff-30d8f6be6320feeacf686be94f48c70869b52630e01ea625f0f15adc0d57c3e4 > was introduced in version 2. > > The hwdata dependency added in v2 might not be needed, I guess it > should be runtime dependency for libfwupdplugin not build time. > > > but now it should have. Then meson.bbclass is setting datadir > > so I'm wondering if there is a bug in meson.bbclass or in these fwupd > > side scripts that they only check the host paths on not the > > recipe sysroot ones? Is fwupd not using the correct variables? > > Yes it checks the paths on host not in RSS as said in commit message. > > > Should the fwupd meson config actually check if the file is in > > join_paths(sys_root, datadir, 'hwdata')? > > Something like this might work, but not sure where you expect sys_root > to be defined. I see it added as a property by meson.bbclass, but > doing just this causes: > ERROR: Unknown variable "sys_root". And if you change the meson.build like this then you would need to set FWUPD_DATADIR_VENDOR_IDS without the sys_root prefix, because in runtime it will be just ${datadir}/hwdata again.
Hi, On Mon, Jul 14, 2025 at 05:13:47PM +0200, Martin Jansa wrote: > On Mon, Jul 14, 2025 at 5:12 PM Martin Jansa <martin.jansa@gmail.com> wrote: > > On Mon, Jul 14, 2025 at 4:31 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > > On Mon, Jul 14, 2025 at 04:00:03PM +0200, Martin Jansa wrote: > > > > On Mon, Jul 14, 2025 at 3:52 PM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > > > > On Mon, Jul 14, 2025 at 03:46:57PM +0200, martin.jansa@gmail.com wrote: > > > > > > From: Martin Jansa <martin.jansa@gmail.com> > > > > > > > > > > > > * it's in RSS as fwupd/2.0.12/recipe-sysroot/usr/share/hwdata/usb.ids but > > > > > > meson is searching the paths on host with: > > > > > > > > > > > > # look for usb.ids in both of the Debian and Fedora locations, > > > > > > # and fall back to the system datadir in case we're building in a venv or prefix > > > > > > vendor_ids_dir = get_option('vendor_ids_dir') > > > > > > if vendor_ids_dir == '' > > > > > > vendor_ids_dir = join_paths(datadir, 'misc') > > > > > > if not fs.is_file(join_paths(vendor_ids_dir, 'usb.ids')) > > > > > > vendor_ids_dir = join_paths(datadir, 'hwdata') > > > > > > > > > > Isn't this the same as ${datadir}/hwdata ? > > > > > > > > It is the same value, but on my host, I don't have > > > > /usr/share/hwdata/usb.ids so meson doesn't detect any existing > > > > vendor_ids_dir value and fails. > > > > > > > > With this change it accepts passed vendor_ids_dir value without > > > > testing it's existence (neither on host nor in RSS). > > > > > > Previous version of fwupd update patches did not include "hwdata" > > > as build dependency so recipe sysroot did not have usb.ids file > > > > and previous version wasn't failing like this, it started to fail only > > after the upgrade as > > https://github.com/fwupd/fwupd/commit/61a863de8dd429be8c393ec267d63db78de71b76#diff-30d8f6be6320feeacf686be94f48c70869b52630e01ea625f0f15adc0d57c3e4 > > was introduced in version 2. > > > > The hwdata dependency added in v2 might not be needed, I guess it > > should be runtime dependency for libfwupdplugin not build time. > > > > > but now it should have. Then meson.bbclass is setting datadir > > > so I'm wondering if there is a bug in meson.bbclass or in these fwupd > > > side scripts that they only check the host paths on not the > > > recipe sysroot ones? Is fwupd not using the correct variables? > > > > Yes it checks the paths on host not in RSS as said in commit message. > > > > > Should the fwupd meson config actually check if the file is in > > > join_paths(sys_root, datadir, 'hwdata')? > > > > Something like this might work, but not sure where you expect sys_root > > to be defined. I see it added as a property by meson.bbclass, but > > doing just this causes: > > ERROR: Unknown variable "sys_root". > > And if you change the meson.build like this then you would need to set > FWUPD_DATADIR_VENDOR_IDS without the sys_root prefix, because in > runtime it will be just ${datadir}/hwdata again. Right, so your patch is the way out for now. Thanks for checking this! Cheers, -Mikko
diff --git a/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb b/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb index e25fbe1f54..eb43c3d865 100644 --- a/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb +++ b/meta-oe/recipes-bsp/fwupd/fwupd_2.0.12.bb @@ -37,6 +37,8 @@ GIDOCGEN_MESON_DISABLE_FLAG = 'disabled' GIR_MESON_ENABLE_FLAG = 'enabled' GIR_MESON_DISABLE_FLAG = 'disabled' +EXTRA_OEMESON = "-Dvendor_ids_dir=${datadir}/hwdata" + PACKAGECONFIG ??= "\ ${@bb.utils.contains('DISTRO_FEATURES', 'systemd', 'systemd', '', d)} \ ${@bb.utils.contains('PTEST_ENABLED', '1', 'tests', '', d)} \