diff --git a/CHANGES.rst b/CHANGES.rst
index 24a1eb62..fcd5e8d2 100644
--- a/CHANGES.rst
+++ b/CHANGES.rst
@@ -3,6 +3,13 @@ hmmlearn Changelog
 
 Here you can see the full list of changes between each hmmlearn release.
 
+next
+----
+
+- Removed the deprecated ``iter_from_X_lengths`` and ``log_mask_zero``;
+  ``lengths`` arrays that do not sum up to the entire array length are no
+  longer supported.
+
 Version 0.2.8
 -------------
 
diff --git a/lib/hmmlearn/_utils.py b/lib/hmmlearn/_utils.py
index cd86782f..5b2ad8c6 100644
--- a/lib/hmmlearn/_utils.py
+++ b/lib/hmmlearn/_utils.py
@@ -1,7 +1,5 @@
 """Private utilities."""
 
-import warnings
-
 import numpy as np
 from sklearn.utils.validation import NotFittedError
 
@@ -12,14 +10,9 @@ def split_X_lengths(X, lengths):
     else:
         cs = np.cumsum(lengths)
         n_samples = len(X)
-        if cs[-1] > n_samples:
+        if cs[-1] != n_samples:
             raise ValueError(
-                f"more than {n_samples} samples in lengths array {lengths}")
-        elif cs[-1] != n_samples:
-            warnings.warn(
-                f"less that {n_samples} samples in lengths array {lengths}; "
-                f"support for silently dropping samples is deprecated and "
-                f"will be removed", DeprecationWarning, stacklevel=3)
+                f"lengths array {lengths} doesn't sum to {n_samples} samples")
         return np.split(X, cs)[:-1]
 
 
diff --git a/lib/hmmlearn/utils.py b/lib/hmmlearn/utils.py
index f8bbb459..a080e561 100644
--- a/lib/hmmlearn/utils.py
+++ b/lib/hmmlearn/utils.py
@@ -1,5 +1,3 @@
-import warnings
-
 import numpy as np
 from scipy import special
 
@@ -57,43 +55,6 @@ def log_normalize(a, axis=None):
         a -= a_lse
 
 
-def iter_from_X_lengths(X, lengths):
-    warnings.warn(
-        "iter_from_X_lengths is deprecated and will be removed in the future.",
-        DeprecationWarning, stacklevel=2)
-    if lengths is None:
-        yield 0, len(X)
-    else:
-        n_samples = X.shape[0]
-        end = np.cumsum(lengths).astype(np.int32)
-        start = end - lengths
-        if end[-1] > n_samples:
-            raise ValueError(
-                f"more than {n_samples} samples in lengths array {lengths}")
-        for i in range(len(lengths)):
-            yield start[i], end[i]
-
-
-def log_mask_zero(a):
-    """
-    Compute the log of input probabilities masking divide by zero in log.
-
-    Notes
-    -----
-    During the M-step of EM-algorithm, very small intermediate start
-    or transition probabilities could be normalized to zero, causing a
-    *RuntimeWarning: divide by zero encountered in log*.
-
-    This function masks this unharmful warning.
-    """
-    warnings.warn(
-        "log_mask_zero is deprecated and will be removed in the future.",
-        DeprecationWarning, stacklevel=2)
-    a = np.asarray(a)
-    with np.errstate(divide="ignore"):
-        return np.log(a)
-
-
 def fill_covars(covars, covariance_type='full', n_components=1, n_features=1):
     if covariance_type == 'full':
         return covars