summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLamont Granquist <lamont@scriptkiddie.org>2018-09-28 12:26:24 -0700
committerLamont Granquist <lamont@scriptkiddie.org>2018-09-28 12:26:24 -0700
commit3ad0741227fc68beb1650b6a9d648126a03a0653 (patch)
treea0e392c5c85fad6c3981c191ea26c69b872b57b0
parent83814a40606d59e2eb4ced35072f0f1859151613 (diff)
downloadchef-3ad0741227fc68beb1650b6a9d648126a03a0653.tar.gz
confusing use case is very confusing. Signed-off-by: Lamont Granquist <lamont@scriptkiddie.org>
-rw-r--r--lib/chef/provider/package/windows.rb8
1 files changed, 7 insertions, 1 deletions
diff --git a/lib/chef/provider/package/windows.rb b/lib/chef/provider/package/windows.rb
index 85e284faa6..cb8b5de151 100644
--- a/lib/chef/provider/package/windows.rb
+++ b/lib/chef/provider/package/windows.rb
@@ -57,7 +57,13 @@ class Chef
# FIXME: this label should not be used. It could be set to nil. Probably what should happen is that
# if the file hasn't been downloaded then load_current_resource must download the file here, and then
# the else clause to set current_resource.version can always be run. Relying on a side-effect here
- # produces at least less readable code, if not outright buggy...
+ # produces at least less readable code, if not outright buggy... (and I'm assuming that this isn't
+ # wholly just a bug -- since if we only need the package_name to determine if its installed then we
+ # need this, so I'm assuming we need to download the file to pull out the name in order to check
+ # the registry -- which it still feels like we get wrong in the sense we're forcing always downloading
+ # and then always installing(?) which violates idempotency -- and I'm having to think way too hard
+ # about this and would need to go surfing around the code to determine what actually happens, probably
+ # in every different package_provider...)
current_resource.version(:unknown.to_s)
else
current_resource.version(package_provider.installed_version)