Django 1.6 GenericRelation admin list_filter regression

[stephenmcd]

Previously in 1.5 it was possible to have a model A with a GenericRelation field to a model B that had a ForeignKey to model C, and use the BC path in A's admin list_filter.

In 1.6 this fails in django.contrib.admin.util.get_model_from_relation where the model on the wrong side of the relationship is returned, when the correct model is available via the parent_model attribute.

I've created a fix and test case here: ​https://github.com/django/django/pull/1913

Although as discussed, the fix should probably be different.

[Anssi Kääriäinen]

A proposed patch for 1.6.x in ​https://github.com/akaariai/django/commit/c895dd39ae4e7687b57ff01ecfcc27caf70323ed.

There are two issues I am concerned with:

• Is the tests in the correct place, and are the inline models the correct way forward?
• Is there more similar regressions in admin? The get_field_by_name() method is used a lot for constructing relations, path names etc. Maybe some of those are affected too.

In general the admin code should be DRYed up a little when it comes to relation path generation.

For master the elif branches in admin/util.py could be removed, it shouldn't be hit at all, but for 1.6 could cause more regressions (shouldn't, but can't assert it doesn't).

[stephenmcd]

Thanks akaariai - I can confirm the above path resolves the issue in Mezzanine.

[Anssi Kääriäinen <akaariai@…>]

In 752d3d70da6291039f33781a0a2ef6f3b7c5fcb5:

Fixed #21431 -- GenRel->FK list_filter regression in admin

Report, analysis and tests from stephenmcd.

[Anssi Kääriäinen <akaariai@…>]

In 5730bf281922151f211a2509dd8b493230d6a187:

[1.6.x] Fixed #21431 -- GenRel->FK list_filter regression in admin

Report, analysis and tests from stephenmcd.

Backpatch of 752d3d70da6291039f33781a0a2ef6f3b7c5fcb5 from master.