| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
Instead of previously having merging which was not backwards compatible
with the 0.7.1 and prior methods this patch works to ensure said backwards
compatible while at the same time making the new merging functionality
work in a more customizable manner.
|
|
|
|
|
|
The new change for merging works well in the mergedict case
but the default merging type for cloud config needs to reflect
how yaml was loaded in bulk, which is the same as the replacing
keys merging type that is now provided.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Looks like this should be in pretty good shape and has passed some
of the basic backwards compat. merging tests that I added.
|
|
|
|
|
|
This patch fixes issues in Fedora 18 (and upcoming RHEL 7) which are
present due to their use of systemd:
- store locale configuration in /etc/locale.conf
- store hostname in /etc/hostname
- use a symlink for /etc/localtime (prior code would set the timezone
but corrupt data in /usr/share/zoneinfo due to presence of symlink)
It also contains fixes for issues unrelated to systemd adoption:
- explicitly scan /dev/sr0 with blkid in order to get the optical drive
in the blkid cache. This prevents an issue on systems running 2.6
kernels (such as RHEL 6) in which config disks on some devices won't
be detected unless the device has previously been queried.
(For reference, see https://patchwork.kernel.org/patch/1770241/)
- append a newline when rewriting sysconfig files, as this is customary
text configuration file formatting and is expected by some parsers
(such as the ifcfg-rh plugin for NetworkManager)
|
|
|
|
a.) appease pylint on raring, as it doesn't like subprocess
pylint: 0.26.0-1ubuntu1
This is mentioned in comments at http://www.logilab.org/ticket/46273
b.) tests/unittests/test_util.py:
the mountinfo lines are longer than 80 chars.
Just disable long lines complaints for this file.
|
|
|
|
1) Refactor util.get_mount_info() to facilitate unit testing.
2) Add unit tests for /proc/$$/mountinfo parsing.
|
|
This re-applies the change in revno 785.
A merge made this test pass rather than skip, but I'd rather have
it skip for now, as we really hope to have the upstart bug fixed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
E1103: 81,44:TestWriteFile.test_basic_usage: Instance of 'Bunch' has no
'st_mode' member (but some types could not be inferred)
so, if it wants st_mode, for now just give it one.
|
|
|
|
|
|
|
|
resize_devices now contains what action occurred for each entry.
|
|
growrun --dry-run will exit 1 if it wouldn't do anything.
so call it, check for '1' and if no change, then just return.
|
|
|
|
|
|
|
|
Now, parser.parse specifies options that override any options found,
rather than just being default options.
There could still potentially be a user for default_options, but since we're
not using them anywhere, I've dropped it. The difference is that in setting up
the root user, we're now insisting that all keys that go in there have the
key_prefix, even if the key content had other options.
I think this is actually the commit that fixes LP: #1136343.
|
|
* drop the parsing of options into csv, as we were only exploding them
back. That can only result in error. Just do minimal parsing.
* change the parsing of key lines to:
if entry is valid:
* use it
else try taking off options:
if good, use it
else fail
|
|
|
|
|
|
|
|
|
|
When the old user: style entry is found, don't forget that
we need to use the distro settings that are provided but
override the name with the new name, this is now accomplished
by merging them together in the correct order (using the standard
cloud-init merging algo).
|
|
Previously if a legacy user: XYZ entry was found, XYZ would not automatically
be promoted to the default user but would instead just be added on as a new
entry to the normalized user list. It appears the behavior that is wanted is
for the XYZ entry to be added on as the default user (thus overriding a distro
provided default user), which better matches how the code previous worked.
LP: #1100920
|
|
previously, there was an attempt in the config drive source to limit
the source device to a "full block device" rather than a partition.
This was done by a simplistic approach of checking that the last
character of the name was not a number. That was filtering out
CD-rom devices (sr0).
Now, we have a bit more sophisticated approach to that same problem.
We filter out block devices that have a 'partition' entry in
/sys/class/block/DEVICE_NAME/partition .
LP: #1100545
|
|
This function can be used to ensure that mocker objects
are restored and verified during usage if exceptions are
thrown while the mock object is being used.
Ensure it is used in the config drive test when multiple
mock objects are being created and restored.
LP: #1098430
|
|
Invoking 'initctl reload-configuration' is only required if inotify
does not work. overlayroot does not support inotify. So, we just
call initctl always, which wont hurt anything.
LP: #1080841
|
