For some reason, the echo|grep pattern was enclosed four times in a $()
operation. Nothing happened, since grep is always invoked with -q
(quiet, nothing printed to stdout).
In the original order, stderr was connected to the old stdout
(terminal). This was _probably_ not intended. The new order fixes this
by first connecting stdout to /dev/null and then attaching stderr to
that as well.
instead of calling the copy command once per file in a directory (eg a
type explorer dir), call the copy command only once with all files of
the directory.
batch copy saves 20% of dry-run time on my test manifest in sequential
mode, and 15% in parallel mode.
Remote::transfer contains a call to mkdir(destination) if the source is
a directory. since the destination must also be a directory we can omit
extra mkdir calls in Explorer.
this saves about 10% on my manifests in sequential mode, and about 6% in
parallel mode.
- Update installation method, following official installation
instructions.
- docker-engine was renamed to Docker CE around March 2017. Update
manpage to reflect that change.
- Remove flag `--experimental` since it is no longer necessary to install a
different binary to get experimental features. Experimental features are
included in the stable binary and can be enabled by a flag or in a
config file.
* add an type explorer to unify detecting of package type.
* update currage use the type explorer, so if os and passed in type
does not match, it behaves correctly.
This type tried to disable an unit after it has removed it, which
failed. Now the removal happens in gencode-remote, after the unit has
been stopped and disabled.
Clarify that this type only operates on units in /etc/systemd/system.
Also, when state=present, it is not always true that the type is
"installed, enabled and started"
Why:
In case when name contains package name with exact version specified (e.g. rpm-build-4.11.3)
```
rpm -q --whatprovides "$pkg_name"
```
will tell you that no package could provide you with 'rpm-build-4.11.3',
because it's not virtual or file-provide, but exact package name.
This will lead to the installation of the package which was already installed.
It slows down manifest execution a lot.
My change will keep previous behaviour which relies on --whatprovides and
will fix wrong behaviour when argument is full package name with version.
This commit adds the following features:
* Ability to expand existing certificate
* Ability to manage object state
* Ability to obtain test certificate
* Ability to promote test certificate to production
* Ability to specify custom certificate name
* Ability to specify multiple domains per certificate
* Ability to use Certbot in standalone mode
* Messaging
This commit also introduces the following behavioral changes:
* Attempt to install Certbot only when it is not installed
already
* Installation of the cron job has to be enabled using
`--automatic-renewal` parameter
**Note:** Object ID is now treated as certificate name and new
required parameter `--domain` was added.
Before this fix, the explorer returned binary data when config did
not exist.
This commit also removes dependency on jq and sets executable flag
on the explorer file.
Error before:
---- BEGIN code-remote:stderr ----
insserv: warning: script 'consul' missing LSB tags and overrides
insserv: There is a loop between service monit and consul if stopped
insserv: loop involving service consul at depth 2
insserv: loop involving service monit at depth 1
insserv: Stopping consul depends on monit and therefore on system facility `$all' which can not be true!
insserv: exiting now without changing boot order!
update-rc.d: error: insserv rejected the script header