search manifests outside of manifests directory

[Dan33l]

The syntax check searches for all .pp files in the modules, which includes
.pp files in the spec tests of the modules like in spec/type_aliases. This can lead to errors like:

bundle exec rake syntax 
---> syntax:manifests
Attempt to redefine entity 'http://puppet.com/2016.1/runtime/type/XXX'. Originally set at file:///builds/XXX/manifests/network.pp?line=24&pos=1.

[Dan33l]

a bugfix was released #100 but reverted #102

[alexjfisher]

@baurmatt I reverted the previous fix after I saw it broke your (and probably many others) workflows. Would you be able to help @Dan33l in finding a solution that works for all?

@Dan33l Could you provide as much detail as possible? Example repo(s) etc so that everybody can reproduce your error?

[baurmatt]

Sure! :) @Dan33l If you can provide examples, I'm happy to take a look!

[Dan33l]

I will provides details as soon as possible.

[rjw1]

I might be missing something but could you not set an exclude path in the config.
PuppetSyntax.exclude_paths = ["vendor/**/*"]

[Dan33l]

I used pdk to build details and reproduce here: https://gitlab.adullact.net/fcombernous/dummy

The pipeline output https://gitlab.adullact.net/fcombernous/dummy/-/jobs/9675
bundle exec rake syntax is enough to fail.

edit : now it works .. but i don't know why. let me search.

edit2 : ah now it fails. since i added a defined type. So perhaps my usage is wrong.

[Dan33l]

well, it looks that the defined type dummy::network and the custom data type Dummy::Network can overlap.

[baurmatt]

Hmm is having a define type and a data type with the same name valid?

[Dan33l]

I used the #office-hours to ask about this. I got the following answer :

I would advise against it. I’m not sure it would cause any errors directly, but it’s best to avoid confusion if possible

[hlindberg]

@baurmatt It is illegal to have a data type with the same name as a resource type (user defined or in ruby) - this because every resource type is also a data type alias.
At runtime, the language handles this by the rule "any referenced non existing data type is a reference to a resource type" and that depends on the order of evaluation.

[Dan33l]

@hlindberg thank you for the answer.

I suppose we can close this issue.