manage.py sqlsequencereset misses m2m tables

[Andrey Golovizin]

Sequence reset code not generated for tables that are implicitly created for m2m fields.

Example models:

class User(models.Model):
    pass

class Group(models.Model):
    users = models.ManyToManyField(User)

The correct result of ./manage.py sqlsequencereset should be:

BEGIN;
SELECT setval('"example_user_id_seq"', coalesce(max("id"), 1), max("id") IS NOT null) FROM "example_user";
SELECT setval('"example_group_id_seq"', coalesce(max("id"), 1), max("id") IS NOT null) FROM "example_group";
SELECT setval('"example_group_users_id_seq"', coalesce(max("id"), 1), max("id") IS NOT null) FROM "example_group_users";
COMMIT;

Actual (incorrect) result:

BEGIN;
SELECT setval('"example_user_id_seq"', coalesce(max("id"), 1), max("id") IS NOT null) FROM "example_user";
SELECT setval('"example_group_id_seq"', coalesce(max("id"), 1), max("id") IS NOT null) FROM "example_group";
COMMIT;

Sequence example_group_users_id_seq is ignored here.

[Andrey Golovizin]

Database backend is postgresql_psycopg2.

[Ramiro Morales]

Bisection between 1.1 (r11366) and current trunk tip (r12889) shows this was introduced in r11710. Accepting and setting target to 1.2 as this is a regression.

[Alex Gaynor]

Almost definitely just a matter of passing auto=True (or whatever the kwarg is named) to cache.get_models()

[Gabriel Hurley]

sets include_auto_created to True for sqlsequencereset

[Gabriel Hurley]

I've attached a patch which causes the auto-created models to be reset as they should, but for the life of me I can't figure out how to write a regression test for this...

If anyone has a bright idea on how to test without comparing SQL line-for-line (which will vary by backend) let me know.

[Russell Keith-Magee]

(In [12903]) Fixed #13248 -- Corrected the output of sqlsequencereset after the m2m changes. Thanks to EroSennin? for the report, and Gabriel Hurley for the fix.

[Jacob]

Milestone 1.2 deleted