diff --git a/.github/workflows/warnings_tests.yaml b/.github/workflows/warnings_tests.yaml
new file mode 100644
index 0000000..cf66876
--- /dev/null
+++ b/.github/workflows/warnings_tests.yaml
@@ -0,0 +1,23 @@
+name: Warnings Tests
+on: [push, pull_request]
+
+jobs:
+  tests:
+    name: Warning Tests
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v3
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.12'
+
+      - name: Install linsolve
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[dev]"
+
+      - name: Run Tests
+        run: |
+          pytest -W error
diff --git a/src/linsolve/linsolve.py b/src/linsolve/linsolve.py
index c16cf8b..550c6c7 100644
--- a/src/linsolve/linsolve.py
+++ b/src/linsolve/linsolve.py
@@ -45,7 +45,7 @@ def ast_getterms(n):
     if type(n) is ast.Name:
         return [[n.id]]
     elif type(n) is ast.Constant or type(n) is ast.Constant:
-        return [[n.n]]
+        return [[n.value]]
     elif type(n) is ast.Expression:
         return ast_getterms(n.body)
     elif type(n) is ast.UnaryOp:
@@ -564,14 +564,20 @@ def _invert_solve(self, A, y, rcond):
         methods.
         """
         # As of numpy 1.8, solve works on stacks of matrices
+        # Change in numpy 2.0:
+        # The b array is only treated as a shape (M,) column vector if it is
+        # exactly 1-dimensional. In all other instances it is treated as a stack
+        # of (M, K) matrices. Previously b would be treated as a stack of (M,)
+        # vectors if b.ndim was equal to a.ndim - 1.
         At = A.transpose([2, 1, 0]).conj()
         AtA = [np.dot(At[k], A[..., k]) for k in range(y.shape[-1])]
-        Aty = [np.dot(At[k], y[..., k]) for k in range(y.shape[-1])]
+        Aty = [np.dot(At[k], y[..., k])[:, None] for k in range(y.shape[-1])]
 
         # This is slower by about 50%: scipy.linalg.solve(AtA, Aty, 'her')
 
         # But this sometimes errors if singular:
-        return np.linalg.solve(AtA, Aty).T
+        print(len(AtA), len(Aty), AtA[0].shape, Aty[0].shape)
+        return np.linalg.solve(AtA, Aty).T[0]
 
     def _invert_solve_sparse(self, xs_ys_vals, y, rcond):
         """Use linalg.solve to solve a fully constrained (non-degenerate) system of eqs.
@@ -690,7 +696,7 @@ def eval(self, sol, keys=None):
     def _chisq(self, sol, data, wgts, evaluator):
         """Internal adaptable chisq calculator."""
         if len(wgts) == 0:
-            sigma2 = {k: 1.0 for k in list(data.keys())}  # equal weights
+            sigma2 = dict.fromkeys(data.keys(), value=1.0)  # equal weights
         else:
             sigma2 = {k: wgts[k] ** -1 for k in list(wgts.keys())}
         evaluated = evaluator(sol, keys=data)
diff --git a/tests/test_linsolve.py b/tests/test_linsolve.py
index 4c32c70..df14d97 100644
--- a/tests/test_linsolve.py
+++ b/tests/test_linsolve.py
@@ -141,7 +141,7 @@ def test_eval(self):
 
 
 class TestLinearSolver:
-    def setup(self):
+    def setup_class(self):
         self.sparse = False
         eqs = ["x+y", "x-y"]
         x, y = 1, 2
@@ -366,7 +366,7 @@ def setup(self):
 
 
 class TestLogProductSolver:
-    def setup(self):
+    def setup_class(self):
         self.sparse = False
 
     def test_init(self):
@@ -466,7 +466,7 @@ def setup(self):
 
 
 class TestLinProductSolver:
-    def setup(self):
+    def setup_class(self):
         self.sparse = False
 
     def test_init(self):
@@ -490,7 +490,9 @@ def test_init(self):
             np.testing.assert_almost_equal(eval(k), 0.002)
         assert len(ls.ls.prms) == 3
 
-        ls = linsolve.LinProductSolver(d, sol0, w, sparse=self.sparse, build_solver=False)
+        ls = linsolve.LinProductSolver(
+            d, sol0, w, sparse=self.sparse, build_solver=False
+        )
         assert not hasattr(ls, "ls")
         assert ls.dtype == np.complex64