Codestin Search App

artirix1927 · 2026-01-15T01:49:47Z

ticket-27489

AI Assistance Disclosure (REQUIRED)

No AI tools were used in preparing this PR.
If AI tools were used, I have disclosed which ones, and fully reviewed and verified their output.

GPT-5 Mini was used to write a verbose prints and warning catching tests, reviewed the output carefully.

Checklist

This PR follows the contribution guidelines.
This PR does not disclose a security vulnerability (see vulnerability reporting).
This PR targets the main branch.
The commit message is written in past tense, mentions the ticket number, and ends with a period.
I have checked the "Has patch" ticket flag in the Trac system.
I have added or updated relevant tests.
I have added or updated relevant docs, including release notes if applicable.
I have attached screenshots in both light and dark modes for any UI changes.

jacobtylerwalls

Thanks, I copied over some findings that I had from working on #20481.

docs/releases/6.1.txt

django/contrib/auth/management/__init__.py

tests/auth_tests/test_management.py

jacobtylerwalls · 2026-02-02T22:27:16Z

tests/auth_tests/test_management.py

+        ct = ContentType.objects.create(app_label="auth_tests", model="oldmodel")
+        actions = ["add", "change", "delete", "view"]
+        for action in actions:
+            Permission.objects.create(
+                codename=f"{action}_oldmodel",
+                name=f"Can {action} old model",
+                content_type=ct,
+            )
+
+        call_command("migrate", "auth_tests", verbosity=0)


I think it would be simpler & model some realism to rely on the behavior of the post_migrate handler, so just migrate one step:

Suggested change

ct = ContentType.objects.create(app_label="auth_tests", model="oldmodel")

actions = ["add", "change", "delete", "view"]

for action in actions:

Permission.objects.create(

codename=f"{action}_oldmodel",

name=f"Can {action} old model",

content_type=ct,

)

call_command("migrate", "auth_tests", verbosity=0)

# Create initial content type and permissions for OldModel.

call_command("migrate", "auth_tests", "0001", verbosity=0)

# Apply the migration that renames OldModel to NewModel.

call_command("migrate", "auth_tests", "0002", verbosity=0)

actions = ContentType._meta.default_permissions

Don't want to force you to do it my way, but after merging the original version Tim Graham observed without comments it was difficult to follow what was going on, so I think keeping the suggested comments here will help.

tests/auth_tests/test_management.py

jacobtylerwalls

Appreciate the updates!

jacobtylerwalls · 2026-02-05T14:54:18Z

tests/auth_tests/test_management.py

+        ct = ContentType.objects.create(app_label="auth_tests", model="oldmodel")
+        actions = ["add", "change", "delete", "view"]
+        for action in actions:
+            Permission.objects.create(
+                codename=f"{action}_oldmodel",
+                name=f"Can {action} old model",
+                content_type=ct,
+            )
+
+        call_command("migrate", "auth_tests", verbosity=0)


Don't want to force you to do it my way, but after merging the original version Tim Graham observed without comments it was difficult to follow what was going on, so I think keeping the suggested comments here will help.

jacobtylerwalls · 2026-02-05T14:55:44Z

django/contrib/auth/management/__init__.py

+                new_name_str = f"Can {action} {verbose_name_raw}"
+
+                if perm.codename == new_codename and perm.name == new_name_str:
+                    continue


This is missing a test.

You think its better to mock the save and test it that way?
Or get the permission and check that the pk says the same which means its same permission (hope that makes sense to you) ?

It's not clear to me whether this condition ever evaluates True. We would need a perm in the database with a codename that __endswith the old codename but also == the new_codename, so that would either happen because of:

substring matches (which I think you said you were looking into)

or MODEL and model both mapping to model

So you could add a test case for a MODEL being renamed to model (seems silly, but might happen on legacy dbs, I guess). Or revisit this depending on how you remove the substring matching.

If it's not reachable, but you want to add a sanity check in the code, I'd be open to adding an assert.

jacobtylerwalls · 2026-02-05T14:56:45Z

django/contrib/auth/management/__init__.py

+
+    for _, from_codename, to_codename, _ in planned:
+        if verbosity >= 2:
+            print(


Also missing a test.

jacobtylerwalls · 2026-02-05T14:57:06Z

django/contrib/auth/management/__init__.py

+            print(
+                style.WARNING(
+                    f"Failed to rename permission {pk} "
+                    f"from '{old}' to '{new}'. "
+                    f"Please resolve the conflict manually."
+                )


We need to capture this output in the test with captured_stdout and assert over it, see test run:

.........Failed to rename permission 463 from 'change_oldmodel' to 'change_newmodel'. Please resolve the conflict manually. .... ---------------------------------------------------------------------- Ran 117 tests in 2.015s

jacobtylerwalls · 2026-02-05T15:01:32Z

django/contrib/auth/management/__init__.py



+def _get_permission_metadata(apps, app_label, model_name):
+    Permission = apps.get_model("auth", "Permission")


We should get the default permissions defined on the model being renamed, not the Permission model itself.

I get them on the model being renamed if its available. If we have lookup error I use the permission model default permissions.

Gotcha, sorry. In that case, I wonder if we should:

move this line into the except

or, just handle LookupError a lot earlier instead of inside this util

create_permissions() returns very early if there's a LookupError.

jacobtylerwalls · 2026-02-05T15:06:50Z

django/contrib/auth/management/__init__.py

+
+        perms = Permission.objects.using(db).filter(
+            content_type__app_label=app_label,
+            codename__endswith=old_suffix,


Like the first version with icontains, I'm still concerned about substring conflation, can we avoid using endswith somehow?

I can just prepare the whole codename and just codename = looking_for_codename. I don't think there is another way to filter permissions other than strings.

jacobtylerwalls · 2026-02-05T16:24:20Z

I think sometimes the GitHub UI isn't super helpful showing replies. I replied to a comment from a previous review: here
#20539 (comment)

In that review, I added a suggestion to replace the manual content type creations with stepwise migrations, e.g. 0001, then 0002:

        # Create initial content type and permissions for OldModel.
        call_command("migrate", "auth_tests", "0001", verbosity=0)
        # Apply the migration that renames OldModel to NewModel.
        call_command("migrate", "auth_tests", "0002", verbosity=0)

You applied that suggestion (thank you!) but without the comments. So, unless you're deeply familiar with the content type internals (I certainly wasn't!) it's not clear that there's a side effect of permissions being created. All I meant to indicated was to keep the comments I put in my suggestion.

You're doing a great job being responsive to all the feedback, this is a tricky area for several reasons and explains why the ticket has been open so long. I'm very positive on the positive of landing this soon 🤝

github-actions · 2026-02-11T00:18:06Z

📊 Coverage Report for Changed Files

-------------
Diff Coverage
Diff: origin/main...HEAD, staged and unstaged changes
-------------
django/contrib/auth/apps.py (100%)
django/contrib/auth/management/__init__.py (98.3%): Missing lines 147
-------------
Total:   62 lines
Missing: 1 line
Coverage: 98%
-------------

Note: Missing lines are warnings only. Some lines may not be covered by SQLite tests as they are database-specific.

For more information about code coverage on pull requests, see the contributing documentation.

artirix1927 · 2026-02-14T03:44:19Z

I think sometimes the GitHub UI isn't super helpful showing replies. I replied to a comment from a previous review: here #20539 (comment)

In that review, I added a suggestion to replace the manual content type creations with stepwise migrations, e.g. 0001, then 0002:
        # Create initial content type and permissions for OldModel.
        call_command("migrate", "auth_tests", "0001", verbosity=0)
        # Apply the migration that renames OldModel to NewModel.
        call_command("migrate", "auth_tests", "0002", verbosity=0)
You applied that suggestion (thank you!) but without the comments. So, unless you're deeply familiar with the content type internals (I certainly wasn't!) it's not clear that there's a side effect of permissions being created. All I meant to indicated was to keep the comments I put in my suggestion.

You're doing a great job being responsive to all the feedback, this is a tricky area for several reasons and explains why the ticket has been open so long. I'm very positive on the positive of landing this soon 🤝

I noticed that my recent PR is failing during the CI/CD stage due to a PostgreSQL connection error.

It also seems that other PRs are experiencing similar failures today, so I wanted to check whether this is a known infrastructure issue or something caused by my changes.

Could you please confirm whether this is on my side or related to the CI setup?

Thank you.

jacobtylerwalls · 2026-02-20T22:16:34Z

Nothing to do with your changes -- there have been a lot of infrastructure problems on CI recently. The GitHub actions have been more resilient than the Jenkins actions, so I would just pay attention to that feedback for now. Sorry.

Changed the pre migrate signal and inserting permission rename operations in the migration plan to a post migrate signal. When having conflicting permission rename, added a note for user with next steps. Avoiding integrity error by finding conflicting perms in advance and skipping them. Added taking permission actions and verbose name raw from model class when available. Added using full literal codename for filtering instead of using substings with endswith/startswith. Hence deleted dead condition for checking if permission is already correct. Added stdout for instead of prints for logging. Brought back respects other db test. Improved some tests by calling migrate 0001 instead of creating permissions manually. Added verbosity prints test. Added warning msg assertion for conflicting permissions test. Updated general docs and 6.1 release docs.

artirix1927 force-pushed the ticket_27489_fix2 branch 4 times, most recently from 92be68e to ef510f8 Compare January 15, 2026 02:19

artirix1927 mentioned this pull request Jan 15, 2026

Fixed #36793, Refs #27489 -- Fixed a test and improved duplicate-permission handling on model renames. #20339

Closed

6 tasks

jacobtylerwalls reviewed Feb 2, 2026

View reviewed changes

artirix1927 force-pushed the ticket_27489_fix2 branch 3 times, most recently from 6531c21 to 176a899 Compare February 3, 2026 23:36

jacobtylerwalls reviewed Feb 5, 2026

View reviewed changes

artirix1927 force-pushed the ticket_27489_fix2 branch 2 times, most recently from e013d89 to d729f81 Compare February 11, 2026 00:07

artirix1927 force-pushed the ticket_27489_fix2 branch 3 times, most recently from 63e5bf8 to 2c74929 Compare February 14, 2026 03:31

artirix1927 force-pushed the ticket_27489_fix2 branch from 2c74929 to 9d4cc95 Compare February 24, 2026 01:39



		def _get_permission_metadata(apps, app_label, model_name):
		Permission = apps.get_model("auth", "Permission")

Uh oh!

Conversation

artirix1927 commented Jan 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

AI Assistance Disclosure (REQUIRED)

Checklist

Uh oh!

jacobtylerwalls left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jacobtylerwalls left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jacobtylerwalls Feb 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jacobtylerwalls commented Feb 5, 2026

Uh oh!

github-actions bot commented Feb 11, 2026

📊 Coverage Report for Changed Files

Uh oh!

artirix1927 commented Feb 14, 2026

Uh oh!

jacobtylerwalls commented Feb 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

artirix1927 commented Jan 15, 2026 •

edited

Loading

jacobtylerwalls Feb 10, 2026 •

edited

Loading