GitLab

apps: Add strip_source_version_suffix support

See merge request smcv/flatdeb!2

This is necessary for packages that were built in OBS which adds a suffix,
but the source isn't kept, and is identical to the base version except
for debian/changelog.
This is already done for runtimes.
The same `strip_source_version_suffix` variable is used for both runtimes
and apps.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Extract giant shell command line used to retrieve source code of the
application and its dependencies, and convert it to Pyhton3.
This change requests that the SDK embeds Pyhton3 to be able to use
`./run.py app` subcommand.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Previously, we lost this as a side-effect of using aptitude to mark
everything non-Essential as "automatically installed". There is no
point in having apt-transport-https but not apt.

In newer versions of apt, where apt-transport-https is a transitional
package, this will have no practical effect.
Signed-off-by: Simon McVittie <smcv@collabora.com>

apps: Allow to use remote repository for runtime

See merge request smcv/flatdeb!5

The remote-url allows to set a remote url instead of a local directory
to use for Platform and SDK runtime.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Fix systemd-nspawn error when running in gitlab CI

See merge request !6

Add dpkg-dev dependency

See merge request !7

Flatdeb uses "dpkg-architecture" that is part of "dpkg-dev"
Signed-off-by: Ludovico de Nittis <ludovico.denittis@collabora.com>

When running in gitlab CI with docker the following error may occur:
 Failed to register machine: Unit machine-root.scope already exists.

We only use systemd-nspawn for temporary containers which are not intended
to be used interactively, so it is not useful for them to be listed by
machined.

Registering with machined can also be harmful: because our root directory
is named `root`, if two concurrent flatdeb instances communicate with the
same machined instance, the second one will try to use the same name as
the first and so will fail to register, causing flatdeb to fail.

This can be avoided by preventing the systemd-nspawn container from
registering with machined (which is supported since systemd v209).
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Remove dependencies to Aptitude and add pre_apt_script

See merge request !3

platformize: Fix systemd-nspawn error when running in gitlab CI

See merge request !4

This allows to perform actions before 'apt action' run.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

prepare-runtime specific code can be done in a pre_apt_script.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

When running in gitlab CI with docker the following error may occur:
platformize | Failed to register machine: Unit machine-root.scope already exists.

We only use systemd-nspawn for temporary containers which are not intended
to be used interactively, so it is not useful for them to be listed by
machined.

Registering with machined can also be harmful: because our root directory
is named `root`, if two concurrent flatdeb instances communicate with the
same machined instance, the second one will try to use the same name as
the first and so will fail to register, causing flatdeb to fail.

This can be avoided by preventing the systemd-nspawn container from
registering with machined (which is supported since systemd v209).
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Fix script execution when bash is not available

See merge request smcv/flatdeb!1

For runtimes where bash is not available put-ld-config-path and
symlink-alternatives fail with error:
put-ldconfig-in-path | sh: 1: /tmp/script/put-ldconfig-in-path: not found

Running those script from outside of chroot fixes it.
Signed-off-by: Frédéric Danis <frederic.danis@collabora.com>

Ubuntu 12.04 somehow has libnss3-dbgsym and libnspr4-dbgsym packages
that are Multi-Arch: no, even though their corresponding libraries are
Multi-Arch: same.
Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

If we have (for example) 400 -dbgsym packages to install, invoking apt
400 times is rather slow due to startup overhead (locking, reading
and updating the database). Instead, invoke apt-cache once to find out
what is available, and use that to compose a single very large apt-get
command-line.

Similarly, we can summarize what was and wasn't installed from the
information we already have, without having to invoke apt a lot.
Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

This can be useful for (derivatives of) older Debian suites like jessie,
which don't have automatic debug symbols.
Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

This reduces the need to have a giant metapackage with all the shared
libraries' detached debug symbols; now we only need to do that for
libraries that are still using legacy debug symbols (foo-dbg).

This also means we tolerate transient uninstallability for debug
symbols: we'll simply produce a runtime without those debug symbols.
This is particularly important when taking binaries from
debian-security, which does not have a corresponding debug symbols
archive (although when security updates get mirrored into
proposed-updates, their detached debug symbols do appear there).

Finally, it also means we don't run into problems with version skew
between debug symbols and the actual libraries, particularly when
adding buster-proposed-updates-debug (which, in general, will be
strictly newer than buster + buster-security) in the hope of getting
detached debug symbols for buster-security.
Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

It's rather out of place there.
Signed-off-by: Simon McVittie <smcv@collabora.com>

This hasn't actually worked since I switched to using debos, and doesn't
seem to have been necessary.
Signed-off-by: Simon McVittie <smcv@collabora.com>

The only Ubuntu precise derivative I'm interested in is now
self-contained, so flatdeb doesn't need to be able to debootstrap
from precise any more.
Signed-off-by: Simon McVittie <smcv@collabora.com>

In some older suites, building packages will fail in a merged-/usr
chroot (#843073). Building packages in a merged-/usr chroot is also,
in general, not guaranteed to result in binaries that work correctly in
an unmerged-/usr environment, so we should probably prefer to build a
more traditional sysroot.
Signed-off-by: Simon McVittie <smcv@collabora.com>

This should speed up packing and rsync'ing of the sysroot and
base tarballs.
Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>

Signed-off-by: Simon McVittie <smcv@collabora.com>