* Remove support for ancient zypper versions
Even SLES11 has zypper 1.x.
* zypper_repository: don't silently ignore repo changes
So far when a repo URL changes this got silently ignored (leading to
incorrect package installations) due to this code:
elif 'already exists. Please use another alias' in stderr:
changed = False
Removing this reveals that we correctly detect that a repo definition
has changes (via repo_subset) but don't indicate this as change but as a
nonexistent repo. This makes us currenlty bail out silently in the above
statement.
To fix this distinguish between non existent and modified repos and
remove the repo first in case of modifications (since there is no force
option in zypper to overwrite it and 'zypper mr' uses different
arguments).
To do this we have to identify a repo by name, alias or url.
* Don't fail on empty values
This unbreaks deleting repositories
* refactor zypper_repository module
* add properties enabled and priority
* allow changing of one property and correctly report changed
* allow overwrite of multiple repositories by alias and URL
* cleanup of unused code and more structuring
* respect enabled option
* make zypper_repository conform to python2.4
* allow repo deletion only by alias
* check for non-existant url field and use alias instead
* remove empty notes and aliases
* add version_added for priority and overwrite_multiple
* add version requirement on zypper and distribution
* zypper 1.0 is enough and exists
* make suse versions note, not requirement
based on comment by @alxgu
The previous version of my regexp did not take into account packages
such as 'p5-Perl-Tidy' or 'p5-Test-Output', so use a greedy match up to
the last occurrance of '-' for matching the package. This regex has
been extensively tested using all packages as provided by pkgsrc-2016Q1[1].
Footnotes:
[1] http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/?only_with_tag=pkgsrc-2016Q1
- make path to pkgin a global and stop passing it around; it's not going
to change while ansible is running
- add support for several new options:
* upgrade
* full_upgrade
* force
* clean
- allow for update_cache to be run in the same task as upgrading/installing
packages instead of needing a separate task for that
Note that since cpanm version 1.6926 its messages are sent to stdout
when previously they were sent to stderr.
Also there is no need to initialize out_cpanm and err_cpanm and
check for their truthiness as module.run_command() and str.find()
take care of that.
* added stdout and stderr outputs
Added stdout and stderr outputs of the results from composer as the current msg output strips \n so very hard to read when debugging
* using stdout for fail_json
using stdout for fail_json so we get the stdout_lines array
The manual check to see if get_bin_path() returned anything is
redundant, because we pass True to the required parameter of
get_bin_path(). This automatically causes the task to fail if the pacman
binary isn't available. Therefore, the code within the if statement
being removed is never called.
dnf: name=PACKAGE state=latest is reponsible for two use cases:
- to install a package if not already installed.
- to update the package to the latest if already installed.
The latter use cases is not handled properly as base.upgrade does not
throw dnf.exceptions.MarkingError if a package is not installed.
Setting base.conf.best = True ensures a package is installed or
updated to the latest when calling base.install.
Sign-off: jsilhan@redhat.com
Sign-off: jchaloup@redhat.com
