RPM Support

The code that did this in 2.7 can be found starting here: https://github.com/pulp/pulp_rpm/blob/2.7-release/plugins/pulp_rpm/plugins/importers/yum/depsolve.py#L315

This case was covered by old unit tests that are currently being skipped. I've retrofit them to work again under mongoengine, and they indicate that this bug is not currently affecting pulp. The right honorable Mr. Bouterse suggested doing a manual end-to-end test covering this case, which I did.

I tried a couple different scenarios with the same result, but the simplest one to set up uses our tried-and-true zoo repo. After syncing the repo to make the package metadata easy to inspect, here's what I found...

Inside the zoo repo are three packages that are relevant to this case. First is walrus, which has two versions available:

> db.units_rpm.find({'name': 'walrus'}, {'name': 1, 'version': 1, '_id': 0})
{ "name" : "walrus", "version" : "0.71" }
{ "name" : "walrus", "version" : "5.21" }

Second is chimpanzee, which requires walrus and squirrel (null values in requires field list removed for brevity):

> db.units_rpm.find({'name': 'chimpanzee'}, {'name': 1, 'requires': 1, '_id': 0})
{ "name" : "chimpanzee", "requires" : [ { "name" : "squirrel" }, { "name" : "walrus" } ] }

Additionally, both versions of walrus are associated with the zoo repo.

> db.units_rpm.find({'name': {$in: ['walrus', 'chimpanzee']}}, {'name': 1})
{ "_id" : "dec1060f-06d3-474c-973b-3caa593d6214", "name" : "chimpanzee" }
{ "_id" : "568462b1-afc0-491f-b780-a251e684ba75", "name" : "walrus" }
{ "_id" : "98936883-fb1d-4ba4-be49-6b593e2811e9", "name" : "walrus" }
> db.repo_content_units.find({'repo_id': 'zoo', 'unit_id': {$in: ['dec1060f-06d3-474c-973b-3caa593d6214', '568462b1-afc0-491f-b780-a251e684ba75', '98936883-fb1d-4ba4-be49-6b593e2811e9']}}).count()
3

If I understand this bug correctly, after creating a new repo and copying chimpanzee over from zoo with dependency solving enabled, both walrus units would be copied. But here's what I see:

(pulp_rpm)[vagrant@dev pulp_rpm]$ pulp-admin rpm repo create --repo-id new                                                                                                                             
Successfully created repository [new]

(pulp_rpm)[vagrant@dev pulp_rpm]$ pulp-admin rpm repo copy rpm --from-repo-id zoo --to-repo-id new --str-eq="name=chimpanzee" --recursive                                                                            
<snip>
Copied:
  camel-0.1-1-noarch
  chimpanzee-0.21-1-noarch
  fox-1.1-2-noarch
  squirrel-0.1-1-noarch
  walrus-5.21-1-noarch

As seen above, chimpanzee also depends on squirrel, which depends on fox and camel. walrus 0.71 depends on whale, but walrus 5.21 does not, so all dependencies are accounted for with only the latest version of each package being copied. Here are chimapanzee's potential dependencies for reference (again, null values removed from the release for brevity):

> db.units_rpm.find({'name': {$in: ['squirrel', 'walrus', 'camel', 'fox']}}, {'name': 1, 'requires': 1, 'version': 1, '_id': 0})
{ "name" : "squirrel", "version" : "0.1", "requires" : [ { "name" : "camel" }, { "name" : "fox" } ] }
{ "name" : "camel", "version" : "0.1", "requires" : [ ] }
{ "name" : "fox", "version" : "1.1", "requires" : [ ] }
{ "name" : "walrus", "version" : "0.71", "requires" : [ { "name" : "whale" } ] }
{ "name" : "walrus", "version" : "5.21", "requires" : [ ] }

I believe that this bug is NOTABUG. The last commit to touch the code in question was a hash that might be familiar to pulp team
members by now, 944a05f7, so I don't have a ready explanation for the behavior that mhrivnak observed.

Here are the steps I took if someone wants to play along in the home game:

pulp-admin rpm repo create --repo-id zoo --feed https://repos.fedorapeople.org/repos/pulp/pulp/demo_repos/zoo/
pulp-admin rpm repo sync run --repo-id zoo
pulp-admin rpm repo create --repo-id new
pulp-admin rpm repo copy rpm --from-repo-id zoo --to-repo-id new --str-eq="name=chimpanzee" 

@smyers, great job testing this! It would appear that this is not a valid bug. I think we should get an ack from mhrivnak and then close it as NOTABUG.

Excellent, I'll follow up with mhrivnak.

In the meantime, I'm going through the tests that I revived and cleaning them up, double-checking their sanity, making sure they're useful, etc. I'll ref this issue when that goes up.

Interesting! I will take a look. This is somewhat surprising, but a happy surprise.

The commit that I thought should have changed the behavior is this one: https://github.com/pulp/pulp_rpm/commit/d97b5e46adc49c8b87c67e1955a5c8204fe94dce

That work happened over a weekend in a bit of a blur, so it's possible I ended up figuring out how to preserve the prior behavior without much trouble, but forgot. :)