diff options
| -rw-r--r-- | lib/chef/provider/package/yum/python_helper.rb | 26 |
1 files changed, 26 insertions, 0 deletions
diff --git a/lib/chef/provider/package/yum/python_helper.rb b/lib/chef/provider/package/yum/python_helper.rb index e595b79278..df46c77c03 100644 --- a/lib/chef/provider/package/yum/python_helper.rb +++ b/lib/chef/provider/package/yum/python_helper.rb @@ -130,6 +130,7 @@ class Chef # name-epoch:version-release # name-epoch:version-release.arch # + # @api private def deconstruct_args(provides) raise "provides must have an epoch in the version to deconstruct" unless provides =~ /^(\S+)-(\d+):(\S+)/ @@ -151,6 +152,31 @@ class Chef end end + # In the default case for the yum provider we now do terrible things with ruby + # to concatenate all the properties together to form a single string to feed to + # the python which favors using returnPackages/searchProvides over the + # searchNevra API. That means that these two different ways of constructing the + # resource are now perfectly identical: + # + # yum_package "zabbix-agent-4.0.15-1.fc31.x86_64" + # + # yum_package "zabbix-agent" do + # version "4.0.15-1.fc31" + # arch "x86_64" + # end + # + # This function handles turning the second form into the first form. + # + # In the case where the epoch is given in the version and we do not have any glob + # patterns that is handled by going the other way and calling deconstruct_args due + # to the yum libraries not supporting that calling pattern other than by searchNevra. + # + # NOTE: This is an ugly hack and should NOT be considered an endorsement of this approach + # towards any kind of features or bugfixes in the DNF provider. I'm doing this + # because YUM is sunsetting at this point and its very difficult to fight with the + # libraries on the python side of things. + # + # @api private def combine_args(provides, version, arch) provides = provides.dup maybe_arch = provides.rpartition(".").last |