diff --git a/src/output.rs b/src/output.rs
index 2cd419b..efee0b3 100644
--- a/src/output.rs
+++ b/src/output.rs
@@ -221,14 +221,16 @@ macro_rules! define_unumeric {
 ///
 /// - only valid UTF-8 data can be printed.
 /// - an `X` format specifier with a `#` flag prints the hex data in uppercase,
-///   but the leading `0x` is still lowercase
+///   but the leading `0x` is still lowercase.
 /// - an `o` format specifier with a `#` flag precedes the number with an `o`
-///   instead of `0`
+///   instead of `0`.
 /// - `g`/`G` (shorted floating point) is aliased to `f`/`F`` (decimal floating
-///   point)
-/// - same for `a`/`A` (hex floating point)
+///   point).
+/// - same for `a`/`A` (hex floating point).
 /// - the `n` format specifier, [`Specifier::WriteBytesWritten`], is not
 ///   implemented and will cause an error if encountered.
+/// - precision is ignored for integral types, instead of specifying the
+///   minimum number of digits.
 pub fn fmt_write(w: &mut impl fmt::Write) -> impl FnMut(Argument) -> c_int + '_ {
     use fmt::Write;
     move |Argument {
diff --git a/tests/tests.rs b/tests/tests.rs
index af74bc5..789a541 100644
--- a/tests/tests.rs
+++ b/tests/tests.rs
@@ -256,3 +256,17 @@ fn test_errors() {
     assert_fmt_err(c"%");
     assert_fmt_err(c"%1");
 }
+
+/// Test that specifying precision with an integral type does _not_
+/// match C. This is one of the documented differences in
+/// `fmt_write`. If this is fixed, make sure to update the
+/// documentation.
+///
+/// https://github.com/lights0123/printf-compat/issues/50
+#[test]
+fn test_int_precision() {
+    unsafe {
+        assert_eq!(c_fmt!(c"%.3d", 1).1, "001");
+        assert_eq!(rust_fmt(c"%.3d".as_ptr(), 1).1, "1");
+    }
+}