Re: rebuilding packages producing no amd64, but other binary-arch packages
On Sat, Jan 12, 2019 at 08:32:42AM +0100, Andreas Beckmann wrote: > Source packages that build binary-arch packages, but no amd64 binary > packages are usually not tested by archive wide rebuilds and may > silently FTBFS for a long time. we build them on amd64/i386/arm64/armhf on https://tests.reproducible-builds.org/debian/ > Does anyone have an idea how to quickly build a list of these for sid? https://tests.reproducible-builds.org/debian/unstable/amd64/index_FTBFS.html lists all our ftbfs cases (replace amd64 with one of the three others), but it's not easy to get the arch:all packages in that list -- cheers, Holger --- holger@(debian|reproducible-builds|layer-acht).org PGP fingerprint: B8BF 5413 7B09 D35C F026 FE9D 091A B856 069A AA1C signature.asc Description: PGP signature
Re: rebuilding packages producing no amd64, but other binary-arch packages
On 2019-01-12 15:26, Holger Levsen wrote: > On Sat, Jan 12, 2019 at 08:32:42AM +0100, Andreas Beckmann wrote: >> Source packages that build binary-arch packages, but no amd64 binary >> packages are usually not tested by archive wide rebuilds and may >> silently FTBFS for a long time. > > we build them on amd64/i386/arm64/armhf on > https://tests.reproducible-builds.org/debian/ Great, so we already have coverage of 4 architectures. >> Does anyone have an idea how to quickly build a list of these for sid? > > https://tests.reproducible-builds.org/debian/unstable/amd64/index_FTBFS.html > lists all our ftbfs cases (replace amd64 with one of the three others), > but it's not easy to get the arch:all packages in that list I don't want only the failures. I want all the source packages building binary packages for a particular architecture ... and then rebuild the ones not covered by reproducible-builds on some porterboxes. Andreas
Re: rebuilding packages producing no amd64, but other binary-arch packages
Hi! On Sat, 2019-01-12 at 08:32:42 +0100, Andreas Beckmann wrote: > I just looked at #917650, the qi FTBFS which goes back to jessie (at least). > > Source packages that build binary-arch packages, but no amd64 binary > packages are usually not tested by archive wide rebuilds and may > silently FTBFS for a long time. > This should not affect packages building only arch:all binary packages, > since they can be (are?) tested during amd64 rebuilds, too. > > Does anyone have an idea how to quickly build a list of these for sid? > Ideally something like > source-package version arch1 arch2 arch3 ... > (where [all amd64] are not in arch-list) You should be able to get this information from the Package-List field from the dsc(5) files, and thus exposed in the Sources files. I don't think we might have yet full coverage in the archive, as not all source packages might have been updated using dpkg-dev >= 1.17.7. Checking now I get 2354 source missing the value, and 27615 with it. I guess for the ones that do not have the arch key in that field you could download the source and run «dpkg-source -b» on it and then check its .dsc? Thanks, Guillem
Re: rebuilding packages producing no amd64, but other binary-arch packages
On 12/01/19 at 08:32 +0100, Andreas Beckmann wrote:
> Hi,
>
> I just looked at #917650, the qi FTBFS which goes back to jessie (at least).
>
> Source packages that build binary-arch packages, but no amd64 binary
> packages are usually not tested by archive wide rebuilds and may
> silently FTBFS for a long time.
> This should not affect packages building only arch:all binary packages,
> since they can be (are?) tested during amd64 rebuilds, too.
>
> Does anyone have an idea how to quickly build a list of these for sid?
> Ideally something like
> source-package version arch1 arch2 arch3 ...
> (where [all amd64] are not in arch-list)
Hi,
Using UDD, with:
select distinct source from packages
where release='sid' and architecture != 'all'
except
select source from packages where release='sid'
and architecture = 'amd64'
order by source;
The complete list is:
adlibtracker2 apex arcboot-installer arm-trusted-firmware atf-allwinner
atitvout cmucl cuse4bsd dgen digitools elilo-installer fdflush fenix
fenix-plugins flash-kernel fnfx freebsd-quota freebsd-smbfs
freebsd-utils fuse4bsd fwupd-arm64-signed fwupd-armhf-signed
fwupdate-arm64-signed fwupdate-armhf-signed fwupdate-i386-signed
fwupd-i386-signed game-data-packager gatos gnumach
google-android-build-tools-installer gpiozero grub-efi-arm64-signed
grub-efi-ia32-signed hurd hurd-libfuse iprutils ixp4xx-microcode k8temp
kfreebsd-10 kfreebsd-defaults kfreebsd-downloader-10
kfreebsd-kernel-headers libcxl libdfp libica libinotify-kqueue libocxl
librtas libservicelog libsoc libsystemd-dummy libvecpf libvpd
linux-signed-arm64 linux-signed-i386 longrun lphdisk lsvpd mac-fdisk
mali-midgard-driver mariadb-connector-c micro-evtd mig mv2120-utils ne10
netdde openssl-ibmca paflib partman-prep partman-ufs partman-zfs pcsx2
pdbg pforth pixbros pixfrogger powerpc-utils ppc64-diag qcontrol qi
raspi3-firmware rpi.gpio s390-dasd s390-netdevice s390-sysconfig-writer
s390-tools s390-zfcp s3switch servicelog smlsharp spellcast steam
steamcmd sysconfig ufsutils xf86-video-omap x-loader
xserver-xorg-video-geode xserver-xorg-video-nv yforth zfsutils
zipl-installer zsnes
(103 rows)
Note that's about source packages that currently have binary packages in
the archive. If they have always failed to build on some architecture,
they wouldn't be listed.
The list with architectures is:
select source, array_agg(distinct architecture)
from packages
where release = 'sid' and architecture != 'all'
and source not in (
select source from packages where release='sid'
and architecture = 'amd64'
)
group by source;
source| array_agg
--+--
adlibtracker2| {i386}
apex | {armel}
arcboot-installer| {mips}
arm-trusted-firmware | {arm64}
atf-allwinner| {arm64}
atitvout | {i386}
cmucl| {i386}
cuse4bsd | {kfreebsd-amd64,kfreebsd-i386}
dgen | {i386}
digitools| {i386}
elilo-installer | {i386}
fdflush | {hurd-i386,i386,mips}
fenix|
{armel,armhf,i386,kfreebsd-i386,mips,mipsel}
fenix-plugins|
{armel,armhf,i386,kfreebsd-i386,mips,mipsel}
flash-kernel | {arm64,armel,armhf}
fnfx | {i386}
freebsd-quota| {kfreebsd-amd64,kfreebsd-i386}
freebsd-smbfs| {kfreebsd-amd64,kfreebsd-i386}
freebsd-utils| {kfreebsd-amd64,kfreebsd-i386}
fuse4bsd | {kfreebsd-amd64,kfreebsd-i386}
fwupd-arm64-signed | {arm64}
fwupd-armhf-signed | {armhf}
fwupdate-arm64-signed| {arm64}
fwupdate-armhf-signed| {armhf}
fwupdate-i386-signed | {i386}
fwupd-i386-signed| {i386}
game-data-packager | {i386}
gatos| {i386}
gnumach | {hurd-i386,i386,kfreebsd-i386}
google-android-build-tools-installer | {i386}
gpiozero | {arm64,armel,armhf}
grub-efi-arm64-signed| {arm64}
grub-efi-ia32-signed | {i386}
hurd | {hurd-i386}
hurd-libfuse | {hurd-i386}
iprutils | {ppc64el}
ixp4xx-microcode | {armel}
k8temp | {kfreebsd-amd64,kfreebsd-i386}
kfreebsd-10 | {kfreebsd-amd64,kfreebsd-i386}
kfreebsd-defaults| {hurd-i386,kfreebsd-amd64,kfreebsd-i386}
k
Re: rebuilding packages producing no amd64, but other binary-arch packages
On 2019-01-14 19:53, Lucas Nussbaum wrote: > On 12/01/19 at 08:32 +0100, Andreas Beckmann wrote: >> Source packages that build binary-arch packages, but no amd64 binary >> packages are usually not tested by archive wide rebuilds and may >> silently FTBFS for a long time. >> Does anyone have an idea how to quickly build a list of these for sid? > Using UDD, with: Many thanks for these queries, I ended up with building package lists with select distinct source, source_version, architecture, release from packages where release = '$DISTRO' and architecture = '$ARCH' and source not in (select source from packages where release='$DISTRO' and architecture = 'amd64'); and rebuilt them for the release architectures on sid, stretch, (jessie) using abel, amdahl, plummer, eller, minkus, zelenka, (and i386 locally) I've also tried hurd-i386/sid, but there does not seem to be a kfreebsd porterbox any more, so that will be skipped. Interestingly there are no packages in the archive that are specific to mips64el (and not also built for amd64). Two new bugs have been reported: fenix-plugins #919340 libdfp #919783 I'll do a buster rebuild later in the freeze. Andreas PS: for the curious, here are my porterbox rebuilding scripts: https://salsa.debian.org/anbe/porterbox/blob/master/bin/porterbox-rebuild-in https://salsa.debian.org/anbe/porterbox/blob/master/bin/porterbox-rebuild-package-list
