diff --git a/core/src/Streamly/Internal/FileSystem/Path/Common.hs b/core/src/Streamly/Internal/FileSystem/Path/Common.hs
index 6f2221db19..f4fb96deda 100644
--- a/core/src/Streamly/Internal/FileSystem/Path/Common.hs
+++ b/core/src/Streamly/Internal/FileSystem/Path/Common.hs
@@ -1337,7 +1337,10 @@ unsafeFromChars :: (Unbox a) =>
     -> Stream Identity Char
     -> Array a
 unsafeFromChars encode s =
-    let n = runIdentity $ Stream.fold Fold.length s
+    -- The encoded array may be longer than the char count. We are encoding it
+    -- twice, but it may still be cheaper than reallocating the array or
+    -- oversizing the array.
+    let n = runIdentity $ Stream.fold Fold.length (encode s)
      in Array.fromPureStreamN n (encode s)
 
 -- XXX Writing a custom fold for parsing a Posix path may be better for
diff --git a/test/Streamly/Test/FileSystem/PosixPath.hs b/test/Streamly/Test/FileSystem/PosixPath.hs
index 6e3c59c0cb..db944c2b6e 100644
--- a/test/Streamly/Test/FileSystem/PosixPath.hs
+++ b/test/Streamly/Test/FileSystem/PosixPath.hs
@@ -48,6 +48,11 @@ testFromString = describe "fromString" $ do
         str (p "/usr/bin") `shouldBe` "/usr/bin"
     it "relative roundtrip" $
         str (p "a/b/c") `shouldBe` "a/b/c"
+    -- test correct array size allocation for unicode encoding
+    it "non-ASCII roundtrip preserves trailing bytes" $
+        str (p "\945.txt") `shouldBe` "\945.txt"
+    it "multi-byte UTF-8 roundtrip (4-byte char)" $
+        str (p "\x1F600/file") `shouldBe` "\x1F600/file"
 
 -------------------------------------------------------------------------------
 -- Separators
diff --git a/test/Streamly/Test/FileSystem/WindowsPath.hs b/test/Streamly/Test/FileSystem/WindowsPath.hs
index fa5cb46596..ad73863cec 100644
--- a/test/Streamly/Test/FileSystem/WindowsPath.hs
+++ b/test/Streamly/Test/FileSystem/WindowsPath.hs
@@ -62,6 +62,13 @@ testFromString = describe "fromString" $ do
         str (p "C:\\Users") `shouldBe` "C:\\Users"
     it "forward slashes preserved on roundtrip" $
         str (p "a/b") `shouldBe` "a/b"
+    -- test correct array size allocation for unicode encoding
+    it "non-ASCII (BMP) roundtrip" $
+        str (p "\945.txt") `shouldBe` "\945.txt"
+    -- Non-BMP chars require a UTF-16 surrogate pair (2 words for 1 char),
+    -- which would be truncated if the array were sized by char count.
+    it "multi-word UTF-16 roundtrip (non-BMP char)" $
+        str (p "\x1F600\\file") `shouldBe` "\x1F600\\file"
 
 -------------------------------------------------------------------------------
 -- Validation