Age | Commit message (Collapse) | Author |
|
ssh runs the command through a shell, so we better quote stuff.
|
|
|
|
Due to the piping, ssh thinks this is interactive and defaults to
lowdelay.
|
|
Given this option, http repositories will be forwarded using a ssh port
forward and rewritten such they work on the remote side forwarding their
traffic through ssh.
|
|
|
|
Now that useradd isn't pulled by apt anymore, it's no longer available
inside the chroot. Use the outer one instead.
|
|
except for debspawn
|
|
|
|
|
|
mypy was failing to infer the type of the dict object and considered it
values Any.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The pwd module started caching /etc/passwd and fails to notice the user
created using useradd.
|
|
|
|
Its future in Python upstream sees uncertain.
|
|
|
|
|
|
|
|
|
|
Doing so bears three advantages:
* Adding more context managers - especially conditional ones - becomes
easier.
* Context managers can be delayed until the point where they're
actually needed without adding to indentation.
* We get rid of two local variables. Less state in your head.
|
|
While mmdebstrap would extract the package precisely to the build_path,
sbuild would create a subdirectory <source>-<upstream_version> inside
the build_path. Changing sbuild to do the mmdebstrap behaviour looks
next to impossible, so this commit changes mmdebstrap to behave like
sbuild.
pbuilder also allows changing the build_path in principle. The
documented BUILDDIR variable works quite similar to the sbuild option
--build-path. pbuilder also allows changing the sub directory using the
undocumented BUILDSUBDIR variable. That's another reason to do it like
sbuild.
|
|
Not escaping a $ would make perl interpolate it at config parsing time.
Fixes: 8ea45852d692 ("sbuild: make environment passing work")
|
|
Fixes: c03690f53941 ("implement basic debspawn backend")
|
|
debusine describes a PackageBuildTask in docs/devel/ontology.rst.
Changes performed herein:
* dscpath -> source_package_path
* dscuri -> source_package_url
* buildarch -> build_architecture
* hostarch -> host_architecture
* options -> build_options
* profiles -> build_profiles
* buildpath -> build_path
|
|
|
|
|
|
Fixes: 29859ef384a5 ("support a new field "parallel"")
|
|
Avoid pylint complaining and we don't need the CompletedProcess object.
|
|
Fixes: 65431c5ef882 ("pbuilder: support running lintian")
|
|
Using the field you can specify an ORed set of positive glob-style
patterns for artifacts to retain in the .output.directory. It defaults
to including all artifacts.
|
|
|
|
Fixes: 29859ef384a5 ("support a new field "parallel"")
|
|
When the wrapped backend fails, streamapi does not produce an artifact
tar on stdout. When this goes missing, we'll see a traceback. Hide that
traceback and adapt the exit code if necessary.
|
|
|
|
|
|
|
|
|
|
Bluntly copied from pbuilder as it does not seem to cause any problems
there.
|
|
If a package is specified by name, it is downloaded inside the chroot
using the embedded deb-src uris. sbuild implements this natively,
pbuilder does not implement this yet, but there is #988419 and
mmdebstrap can make it work.
|
|
Instead of writing a complex shell script in a limited sub-language of
shell, call out to ourselves and implement the build functionality in
Python.
|
|
|
|
Fixes: a5d4cfd4b996 ("mmdebstrap: don't keep downloaded .deb files")
|
|
Since we delete the chroot anyway, there is no need to keep the .debs
for the build.
|
|
By doing the unpack before the satisfying build dependencies, we can
point apt at the extracted source package and stop talking about the
.dsc file. It also means that we can clean those artifacts earlier.
|