Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

C:\bug>python
Python 3.3.5rc2 (v3.3.5rc2:ca5635efe090, Mar  2 2014, 18:18:29) [MSC v.1600 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> exit()

C:\bug>python test2.py
File "test2.py", line 1
SyntaxError: encoding problem: iso-8859-1

It's a duplicate of the issue bpo-20731.

It seems that this is not fixed in 3.3.5. Someone please reproduce it.

Works fine for me

Thanks Mark.

Perhaps, the problem is text-mode handling. When using Windows's text-mode stream, ftell() may return -1 even if no error occured.

When opening LF-newline file, ftell() may return zero when the position is not at the beginning of the file.

Maybe LF-newline file should open in binary-mode.
http://support.microsoft.com/kb/68337

I can reproduce this one. There are a few conditions which needs to be met:

• Linux line endings
• File needs to have at least x lines (empty lines are fine). I guess this is the point why no one could reproduce it. The attached file has 19 lines but probably no one copy/pasted the empty lines. Downloading the file reproduces this in my case. The length of the encoding declaration is relevant to the number of required newlines. #coding:latin-1 fails at a file with 19 lines, #coding: latin-1 (whitespace added) requires 20 lines.

More observations:

• Also reproducible if utf8 is used as alias for utf-8 (#coding: utf8 + 17 lines), but not reproducible with utf-8
• Python 3.4 is affected, too
• No issues on Python 3.3.2

I can reproduce this with 3.4.1 and 3.5.0.

This fix for bpo-20731 doesn't address this bug completely because it's possible for ftell to return -1 without an actual error, as test2.py demonstrates.

In text mode, CRLF is translated to LF by the CRT's _read function (Win32 ReadFile). So the buffer that's used by FILE streams is already translated. To get the stream position, ftell first calls _lseek (Win32 SetFilePointer) to get the file pointer. Then it adjusts the file pointer for the unwritten/unread bytes in the buffer. The problem for reading is how to tell whether or not LF in the buffer was translated from CRLF? The chosen 'solution' is to just assume CRLF.

The example file test2.py is 33 bytes. At the time fp_setreadl calls ftell(tok->fp), the file pointer is 33, and Py_UniversalNewlineFgets has read the stream up to '#coding:latin-1\n'. That leaves 17 newline characters buffered. As stated above, ftell assumes CRLF, so it calculates the stream position as 33 - (17 * 2) == -1. That happens to be the value returned for an error, but who's checking? In this case, errno is 0 instead of the documented errno constants EBADF or EINVAL.

Here's an example in 2.7.7, since it uses FILE streams:

    >>> f = open('test2.py')
    >>> f.read(16)
    '#coding:latin-1\n'
    >>> f.tell()
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    IOError: [Errno 0] Error

Can the file be opened in binary mode in Modules/main.c? Currently it's using _Py_wfopen(filename, L"r"). But decoding_fgets calls Py_UniversalNewlineFgets, which expects binary mode anyway.

I've tried to make the title more meaningful, feel free to change it if you can think of something better.

This bug just bit me. Changing "# coding: utf8" to "# coding: utf-8" works around it.

(oops: with Python 3.4.1 on Windows)

I've just been bitten by this on 3.6.2, Windows Server 2008 R2, when running the setup.py script for QuantLib-SWIG:
https://github.com/lballabio/QuantLib-SWIG/blob/v1.10.x/Python/setup.py

It seems there is different behaviour depending on whether:

• Unix (LF) or Windows (CRLF) line endings are used
• The file is >4096 bytes or <=4096 bytes
• The module docstring has an initial space

Some of that has been mentioned previously, but I think the 4096-byte limit might be new, which is why I'm posting.

I've attached a script I used to come up with the results below. It contains:

• a -*- coding line (for iso-8859-1 in this case)
• a docstring consisting entirely of lines of x's, of length 78
• Unix line endings

The file's length is exactly 4096 bytes.

Running this, or slightly modified versions of this, with a 3.6.2 interpreter gave the following results:

• In all cases, when Windows line endings were used there was no issue - running the script produced no errors or output.

• With Unix line endings:

  • File length <= 4096, with no leading spaces in the docstring:
    File "bpo-20844.py", line 1
    SyntaxError: encoding problem: iso-8859-1

  • File length > 4096, with no leading spaces in the docstring:
    File "bpo-20844.py", line 56
    xxxxx"""
    ^
    SyntaxError: EOF while scanning triple-quoted string literal

  • Any file length, with the first 'x' on line 3 replaced with a space (line 2 if the coding line is ignored):
    File "bpo-20844.py", line 2
    xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
    ^
    IndentationError: unexpected indent

I had no issues with python 2.7.13.

New changeset 10654c1 by Inada Naoki in branch 'master':
bpo-20844: open script file with "rb" mode (GH-12616)
10654c1

New changeset 8384670 by Inada Naoki in branch '3.7':
bpo-20844: open script file with "rb" mode (GH-12616)
8384670