OSError: [Errno 36] File name too long in Python (tested)

OSError: [Errno 36] File name too long (ENAMETOOLONG) means one path component is over 255 bytes or the whole path is over 4096 bytes on Linux. Non-ASCII names hit the limit sooner because UTF-8 uses several bytes per character. Shorten the name by bytes and add a hash to keep it unique.

1. The error

name = "report_" + "x" * 250 + ".csv"
print(len(name))
open(name, "w")

Output:

261
Traceback (most recent call last):
  File "/tmp/ln372/b1.py", line 3, in <module>
    open(name, "w")
    ~~~~^^^^^^^^^^^
OSError: [Errno 36] File name too long: 'report_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.csv'

The file name is 261 characters. Linux filesystems such as ext4 allow at most 255 bytes per path component (NAME_MAX), so open() fails with ENAMETOOLONG. This usually happens when a file name is built from user input, a URL, a page title or a long list of parameters.

2. The limit is 255 bytes, not 255 characters

import os
print(os.pathconf(".", "PC_NAME_MAX"), os.pathconf(".", "PC_PATH_MAX"))
ok = "a" * 255
open(ok, "w").close()
print("255 bytes: ok")
os.remove(ok)
# 128 characters, but each "é" is 2 bytes in UTF-8
name = "é" * 128
print(len(name), "chars,", len(name.encode()), "bytes")
try:
    open(name, "w")
except OSError as e:
    print(type(e).__name__, e.errno, e.strerror)

Output:

255 4096
255 bytes: ok
128 chars, 256 bytes
OSError 36 File name too long

Exactly 255 ASCII bytes works. But 128 characters of é is 256 bytes in UTF-8 and fails, so a len(name) <= 255 check is not enough for non-ASCII names (accents, CJK, emoji). Check len(name.encode()) instead. os.pathconf(".", "PC_NAME_MAX") reports the limit for the filesystem you are on.

3. The whole path is too long (PATH_MAX 4096)

import os
# each component is only 200 bytes, but the whole path is > 4096 (PATH_MAX)
path = "/".join(["d" * 200] * 21) + "/f.txt"
print(len(path))
try:
    os.makedirs(os.path.dirname(path))
except OSError as e:
    print(type(e).__name__, e.errno, e.strerror, "| len(filename) =", len(e.filename))
depth = 0
p = "d" * 200
while os.path.isdir(p):
    depth += 1
    p = os.path.join(p, "d" * 200)
print("directories created before the failure:", depth)

Output:

4226
OSError 36 File name too long | len(filename) = 4220
directories created before the failure: 20

Each component here is only 200 bytes, but the full path passed to the kernel may not exceed 4096 bytes (PC_PATH_MAX). os.makedirs created 20 levels and then failed on the 21st, so it leaves a partial tree behind. Deeply nested generated directories (cache keys, mirrored URLs) are the usual cause.

4. Fix: shorten the name by bytes and add a hash

import hashlib, os

def safe_filename(name, max_bytes=255):
    """Shorten name to fit NAME_MAX bytes, keeping the extension and a hash."""
    if len(name.encode()) <= max_bytes:
        return name
    stem, ext = os.path.splitext(name)
    digest = hashlib.sha1(name.encode()).hexdigest()[:10]
    budget = max_bytes - len(ext.encode()) - len(digest) - 1
    short = stem.encode()[:budget].decode("utf-8", "ignore")
    return f"{short}-{digest}{ext}"

for n in ["report.csv", "report_" + "x" * 250 + ".csv", "é" * 200 + ".txt"]:
    s = safe_filename(n)
    print(len(n.encode()), "->", len(s.encode()), s[-20:])
    open(s, "w").close(); os.remove(s)

Output:

10 -> 10 report.csv
261 -> 255 xxxxx-5c000bdbbb.csv
404 -> 255 ééééé-c8a0a30f55.txt

Truncate the UTF-8 bytes, decode with "ignore" so a multi-byte character is not cut in half, keep the extension, and append a short hash of the full name so two long names that share a prefix do not collide. If you need the original name, store it in a sidecar file or database instead of in the file name.

5. Catch it precisely

import errno
try:
    open("y" * 300, "w")
except OSError as e:
    print(type(e).__name__, e.errno == errno.ENAMETOOLONG, errno.errorcode[e.errno])

Output:

OSError True ENAMETOOLONG

It is a plain OSError with no dedicated subclass; check e.errno == errno.ENAMETOOLONG.

Related: [Errno 22] Invalid argument · [Errno 2] No such file or directory · [Errno 39] Directory not empty.

Written with AI assistance. Measured on 2026-10-23 with Python 3.13.5 on Linux (Debian 13, ext4). Each block was run verbatim and the output pasted unedited. PlainTables is our extension; Python is free.