Replace usage of _pulp_client with pubtools-pulplib [RHELDST-141]

[negillett]

This commit removes the library's native _pulp_client ("legacy") and replaces it with usage of pubtools-pulplib methods for associating and unassociating content.

Additionally, PulpActions were replaced with simple abstractions for grouping content to be associated/unassociated.

Tests were removed/edited as necessary.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.78%. Comparing base (13ab604) to head (8455d0c).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #223      +/-   ##
==========================================
- Coverage   98.85%   98.78%   -0.08%     
==========================================
  Files           8        7       -1     
  Lines         960      820     -140     
==========================================
- Hits          949      810     -139     
+ Misses         11       10       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

I'm at a total loss as to why we're seeing error 111 on '/api/v1/manifest'. I've touched nothing in the ubi client.
Any ideas, @rbikar?

[rbikar]

run tests

[rbikar]

I'm at a total loss as to why we're seeing error 111 on '/api/v1/manifest'. I've touched nothing in the ubi client. Any ideas, @rbikar?

I think the ubi-manifest service in the test env was in a bad shape, I re-ran tests and they passed OK.

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

run tests

[negillett]

@rbikar, @rohanpm, this is ready for review.
I'm still unsure why some modulemd profiles no longer contain "[d]" to signal default but I don't believe the change was introduced here (it was only ever present in the integration tests). It could be a difference between rhel7 and rhel8 content--this is a key difference between the two integration tests that asserted the presence of the mark.

[rbikar]

@rbikar, @rohanpm, this is ready for review. I'm still unsure why some modulemd profiles no longer contain "[d]" to signal default but I don't believe the change was introduced here (it was only ever present in the integration tests). It could be a difference between rhel7 and rhel8 content--this is a key difference between the two integration tests that asserted the presence of the mark.

In realistic usage the ubi7 shouldn't have any modules (but I guess in test the situation can be different).

Anyway you need to examine modulemd_defaults units in the test instance of rhsm-pulp before and after the test.

I checked current real ubi8 image and e.g. for perl module I get this:
Output of dnf module list

perl                 5.24       common [d Practical Extraction and Report Language                                                                                                                                                             
                                ], minima 
                                l         
perl                 5.26 [d]   common [d Practical Extraction and Report Language                                                                                                                                                             
                                ], minima 
                                l         
perl                 5.30       common [d Practical Extraction and Report Language                                                                                                                                                             
                                ], minima 
                                l         
perl                 5.32       common [d Practical Extraction and Report Language                                                                                                                                                             
                                ], minima 
                                l   

and corresponding modulemd_defaults unit in production rhsm-pulp data:
(Maybe the testing code reads wrongly the output as in terminal the formatting doesn't look very nice)

...
            "name": "perl",
            "profiles": {
                "5.24": [
                    "common"
                ],
                "5.26": [
                    "common"
                ],
                "5.30": [
                    "common"
                ],
                "5.32": [
                    "common"
                ]
            },
            "pulp_user_metadata": {},
            "repo_id": "ubi-8-for-x86_64-appstream-rpms__8",
            "stream": "5.26"
        },
...

which say 2 things:

• what profile is default for which stream
• what stream is default for the module

so it should be dependent on the data in pulp only, can you double-check the removal and copy of modulemd_defaults units? There might some subtle bug (or the formatting problem as I mentioned above).

[negillett]

run tests

[negillett]

run tests

[negillett]

"common [d]" is showing up now for that one tests 🤷🏻

[negillett]

run tests

[negillett]

@rohanpm @rbikar ptal