Fix support for null fullname in orbit db

[charliemirabile]

The code that handles the fullname field is designed with the idea that it can be null (i.e. warpdrive does not enforce it being provided, and the containerfile code checks if it exists instead of using it straight up), but the actual db schema does not include null=True. Correct this so that accounts not intended for use by students can be created without needing to provide a name

[krishjainx]

Would this patch be good enough to resolve this issue?

diff --git a/orbit/db.py b/orbit/db.py  
index ded2d8d..a1f88c1 100755  
--- a/orbit/db.py  
+++ b/orbit/db.py  
@@ -15,7 +15,7 @@ class User(BaseModel):  
     username = peewee.TextField(unique=True)  
     pwdhash = peewee.TextField(null=True)  
     student_id = peewee.TextField(unique=True, null=True)  
-    fullname = peewee.TextField()  
+    fullname = peewee.TextField(null=True)  

This fixes the issue where fullname was required even though the rest of the codebase assumes it can be NULL.

One limitation is that if the test script is run twice, it will still hit a UNIQUE constraint failed: user.student_id error, since test.sh tries to create a user with student_id = 1234, which already exists from the first run. This happens because the database persists between runs in the Podman volume. Clearing the volume before running the tests would prevent duplicate users from previous runs.

That said, this change should at least resolve the NOT NULL constraint failed: user.fullname issue!

---

Currently, running test.sh requires pruning volumes beforehand, which isn’t ideal and should be fixed as part of a broader testing overhaul.

Here’s the script Joel uses for a full testing cycle:

#!/bin/sh
set -ex

podman-compose down || true # Not a big deal if this fails since they might not be up
yes | podman volume prune # Slightly scary in case you have other volumes you care about. FIXME: could name them explicitly
rm -rf socks
podman-compose build
podman-compose up -d
# Wait until Synapse is done initializing
podman-compose logs -f submatrix 2>&1 | sed '/Synapse now listening on TCP port 8008/ q'
./test.sh
podman-compose down

Next Steps:

I see two possible approaches here:

A) I can modify the script to wrap test.sh execution appropriately—moving its steps earlier in the script while ensuring podman-compose down is properly placed at the end. Then, I’ll make a PR for that.

B) Alternatively, if you'd prefer a more comprehensive fix, I can read through more of the code and work on a proper testing overhaul to eliminate the need for volume pruning beforehand. If that sounds good, I can run my thoughts by you before making a PR

Let me know which approach you’d prefer!

[charliemirabile]

The patch looks fine. I am not sure that we need to do anything about the tests since really we only had this regression because I pushed the change to add the full names without running the tests first.

[krishjainx]

@charliemirabile Great! Should I modify test.sh to prune volumes beforehand? Running the script twice causes a UNIQUE constraint failed error since the database persists in the Podman volume. Clearing it before tests would prevent duplicate user issues. Let me know if you'd like a PR for this! (along with this patch)

diff --git a/orbit/db.py b/orbit/db.py  
index ded2d8d..a1f88c1 100755  
--- a/orbit/db.py  
+++ b/orbit/db.py  
@@ -15,7 +15,7 @@ class User(BaseModel):  
     username = peewee.TextField(unique=True)  
     pwdhash = peewee.TextField(null=True)  
     student_id = peewee.TextField(unique=True, null=True)  
-    fullname = peewee.TextField()  
+    fullname = peewee.TextField(null=True)  

[theyoyojo]

no, don't modify test.sh

Other than that your changes looks good.

Time to PR