We are at the point that all distributions we are building have Python
3, so any tools running in the chroot can assume Python 3 exists.
This makes dib-python redundant; mark it as deprecated and start to
remove it from elements where it is no longer required.
Change-Id: I5d852843ec65d3b04444b77c54c5b82424455cd8
Now that DIB is python3 only we can remove a hack that made sure
scripts outside the chroot ran with the correct version of python.
This is necessary as python3 does not resolve symbolic links to the
binary like python2.x did, which causes element scripts to fail finding
modules when DIB was run from inside a venv.
This patch does the following:
1. Reverts 9c7b8d1714 which was the
workaround for mixed python2/3 environments.
2. Updates the scripts to use "python3" instead of "python".
Change-Id: If2402bb02fc8a4778fa9434fa167ea1fafd87c28
The change Ia6f10741fa6be24b11d6991c8a6b6e07951ff68d introduced having
"when:" as a list of values. However, this was actually not
sufficient to express the logic required for arm64/x86_64/xenial
kernel matching we wanted.
Because the package name is a key, we can't have multiple entires in
the package-map YAML files. This means we can't do more advanced
matching and thus we need to be able to match through multiple
parameters. Similar to Ia6f10741fa6be24b11d6991c8a6b6e07951ff68d we
modify the matching rules to allow a list.
A an example of using this is provided in the README.rst, and this
same example worked through by the unit tests.
This also slightly updates the matching logic to be more sequential.
After each check we either continue on or log the failure and continue
to the next check (rather than set a list of flags then check that at
the end). This makes it much easier to understand what is being
matched in the logging output from the tool.
Change-Id: Idff7b067ad4255e6fc4138f7eff313a81b75c8ba
Allow the "when:" statements to be a list of values, which are
effectively anded together to filter the package install.
Change-Id: Ia6f10741fa6be24b11d6991c8a6b6e07951ff68d
Sometimes an element needs packages installed so that it can
perform tasks but those package are not appropriate for the
final image content. Add a "build-only" flag to package-install-squash
which will cause package to be installed at the beginning of the
phase and then uninstalled at the end of the phase.
Change-Id: Ie01b795991710c93f6b669c8f14b57eb4412c1d5
This causes problems for other projects incorporating dib; we don't
have a specific need for a cap.
Fix a few issues, mostly spacing or regex matches. No functional
changes.
W503 and W504 relate to leaving artithmetic operators at the start or
end of lines, and are mutually exclusive and, due to "ignore"
overriding the defaults both get enabled. It seems everyone gets this
wrong (https://gitlab.com/pycqa/flake8/issues/466). Don't take a
position on this and ignore both.
Use double # around comments including YAML snippets using "# type: "
which now gets detected as PEP484/mypy type hints.
Change-Id: I8b7ce6dee02dcce31c82427a2441c931d136ef57
Provide a "when" option that provides for not installing packages
based on a = or != match on an environment variable.
Unit tests are added.
Change-Id: Ifa824dccaff69fd447f45d54cb4a3083bcabdd86
When using the package-installs element there can be some encoding
problems if the package installation emits unparsable output
[1]. However in this case we just want to forward the output to the
console which normally can handle this correctly. In order to fix this
switch off universal_newlines processing such that we just operate on
bytes.
Further we have to decode the lines without setting the locale and
ignoring errors. This is required because print encodes without
setting the locale and thus we need to filter/modify the stream such
that it doesn't crash.
[1] Traceback:
2018-03-01 09:58:00.515 | Traceback (most recent call last):
2018-03-01 09:58:00.515 | File "/usr/local/bin/package-installs-v2", line 137, in <module>
2018-03-01 09:58:00.515 | main()
2018-03-01 09:58:00.515 | File "/usr/local/bin/package-installs-v2", line 130, in main
2018-03-01 09:58:00.515 | process_output(install_args, follow=True)
2018-03-01 09:58:00.515 | for line in iter(proc.stdout.readline, ''):
2018-03-01 09:58:00.515 | File "/usr/lib/python3.5/encodings/ascii.py", line 26, in decode
2018-03-01 09:58:00.515 | return codecs.ascii_decode(input, self.errors)[0]
2018-03-01 09:58:00.515 | UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 34: ordinal not in range(128)
Change-Id: Ie4af9b4523459a630cfb98d09093bfe9ef7aa61e
The current output for package-installs-v2 is inscrutable [1]
The problem starts with process_output() which is not capturing
stderr. This means that any stderr output is dislocated from any
stdout output around it. This is *really* confusing as you get a
bunch of seemingly meaningless stderr output from any calls before you
see any stdout (e.g. in [1] you can see random yum error output that
should have been with the yum call)). The simplest thing to do is to
redirect stderr to stdout which keeps everything in sync.
This causes a slight problem, however, because pkg-map outputs both
status information and errors on stderr. To work around this but
maintain compatibility, we add a "--prefix" argument that prepends
mapped packages from pkg-map with a value we can match on. The
existing status/debug output from pkg-map is low-value; modify the
call so that it will be traced only at higher debug levels (e.g. -x
-x).
The current loop is also calling pkg-map for every package in every
element (this is why in [1] the same message is repeated over and
over). This is unnecessary; it only needs to pkg-map once for each
element, giving the package list as the arguments. Create package
lists by element and pass those to pkg-map.
As a cleanup, there is no point in printing e.output if the
process_output fails for the install because we are already tracing
it; i.e. the output, even for failures, is already in the logs.
Printing it again just duplicates the output.
[2] is an extract showing what I feel is a much more understandable
log output for a fairly complex install.
[1] http://paste.openstack.org/show/595118/
[2] http://paste.openstack.org/show/595303/
Change-Id: Ia74602a5d2db032a476481caec0e45dab013d54f
Currently we have all our elements and library files in a top-level
directory and install them into
<root>/share/diskimage-builder/[elements|lib] (where root is either /
or the root of a virtualenv).
The problem with this is that editable/development installs (pip -e)
do *not* install data_files. Thus we have no canonical location to
look for elements -- leading to the various odd things we do such as a
whole bunch of guessing at the top of disk-image-create and having a
special test-loader in tests/test_elements.py so we can run python
unit tests on those elements that have it.
data_files is really the wrong thing to use for what are essentially
assets of the program. data_files install works well for things like
config-files, init.d files or dropping documentation files.
By moving the elements under the diskimage_builder package, we always
know where they are relative to where we import from. In fact,
pkg_resources has an api for this which we wrap in the new
diskimage_builder/paths.py helper [1].
We use this helper to find the correct path in the couple of places we
need to find the base-elements dir, and for the paths to import the
library shell functions.
Elements such as svc-map and pkg-map include python unit-tests, which
we do not need tests/test_elements.py to special-case load any more.
They just get found automatically by the normal subunit loader.
I have a follow-on change (I69ca3d26fede0506a6353c077c69f735c8d84d28)
to move disk-image-create to a regular python entry-point.
Unfortunately, this has to move to work with setuptools. You'd think
a symlink under diskimage_builder/[elements|lib] would work, but it
doesn't.
[1] this API handles stuff like getting files out of .zip archive
modules, which we don't do. Essentially for us it's returning
__file__.
Change-Id: I5e3e3c97f385b1a4ff2031a161a55b231895df5b
