# Shutil.copyfile Corrupts Copied File When Running Under pythonw on Windows 10 Home

**URL:** <https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019>\
**Category:** Python Help\
**Created:** [July 12, 2024, 5:01pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019 "2024-07-12T17:01:22Z")\
**Posts on this page:** 11\
**Page:** 1

<div class="post-metadata">

**Author:** ![BoQsc](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/boqsc/32/20296_2.png) [@BoQsc](https://discuss.python.org/u/BoQsc)\
**Post date:** [July 12, 2024, 5:01pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/1 "2024-07-12T17:01:22Z")

</div>

I encountered an issue where `shutil.copyfile` corrupts the copied file when executed under `pythonw.exe` on Windows 10 Home. However, running the same operation with `python.exe` works correctly.

#### Issue Description

- **Problem:** When using `shutil.copyfile` in a `.pyw` script executed with `pythonw.exe`, the copied file becomes corrupted.
- **Observation:** Running the same `.pyw` script with `python.exe` results in a correctly copied file.

#### Workaround

- **Solution:** Execute the `shutil.copyfile` operation using `python.exe` instead of `pythonw.exe`.

#### Test Case

To reproduce the issue, use the following script:

### `program.pyw`

```python
import os, subprocess, shutil, sys, tempfile
os.makedirs(os.path.join(tempfile.gettempdir(), "modifield_python"), exist_ok=True)
renamed_pythonexefile = shutil.copyfile(sys.executable, os.path.join(tempfile.gettempdir(), "modifield_python\\modified_python.exe"))

```

For demonstration, this script creates a temporary directory named `modifield_python` and copies the current `Python.exe` executable, renaming to `modified_python.exe`.

### Testing:

Upon launching the by double clicking `C:\Users\%USERNAME%\AppData\Local\Temp\modifield_python\modified_python.exe`, you will sense that something is not right already: it does not open. The size of the copied executable seems reasonable, but it just doesn’t work anymore.

---

<div class="post-metadata">

**Author:** ![barry-scott](https://avatars.discourse-cdn.com/v4/letter/b/e9c0ed/32.png) [@barry-scott](https://discuss.python.org/u/barry-scott)\
**Post date:** [July 12, 2024, 5:07pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/2 "2024-07-12T17:07:51Z")

</div>

I suggest you compare the contents of the two files.  
I expect you will find they are identical.

The issue is likely to be that a windows gui .exe needs additional meta data files for it to launch, which you did not copy.

---

<div class="post-metadata">

**Author:** ![BoQsc](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/boqsc/32/20296_2.png) [@BoQsc](https://discuss.python.org/u/BoQsc)\
**Post date:** [July 12, 2024, 5:19pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/3 "2024-07-12T17:19:00Z")

</div>

Yeah I suspected it might be the case. However they are not completely identical as you would expect.

 ![image](https://us1.discourse-cdn.com/flex002/uploads/python1/original/3X/5/b/5bf83c34d650115336457ddd6e3bc14e6a15a177.png)

Later on I will try to use `shutil.copy2` and see if it copied file will be functional.

> To preserve all file metadata from the original, use [`copy2()`](https://docs.python.org/3/library/shutil.html#shutil.copy2) instead.

* * *

- [shutil — High-level file operations — Python 3.12.4 documentation](https://docs.python.org/3/library/shutil.html#shutil.copy2:~:text=To%20preserve%20all%20file%20metadata%20from%20the%20original%2C%20use%20copy2()%20instead).

---

<div class="post-metadata">

**Author:** ![MegaIng](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/megaing/32/16162_2.png) [@MegaIng](https://discuss.python.org/u/MegaIng)\
**Post date:** [July 12, 2024, 5:27pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/4 "2024-07-12T17:27:13Z")

</div>

> [@BoQsc](#):
>
> The size of the copied executable seems reasonable, but it just doesn’t work anymore.

They are not identical, so no, it does not seem reasonable.

---

<div class="post-metadata">

**Author:** ![BoQsc](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/boqsc/32/20296_2.png) [@BoQsc](https://discuss.python.org/u/BoQsc)\
**Post date:** [July 12, 2024, 6:07pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/5 "2024-07-12T18:07:04Z")

</div>

### [`shutil.copy2()`](https://docs.python.org/3/library/shutil.html#shutil.copy2) worked.

By simply replacing `shutil.copyfile()` with `shutil.copy2()` it all started to work.  
Mostly likely it was due to metadata not being copied.

The size of copied file might not be as original, but it’s still identical to `shutil.copyfile()`

 ![image](https://us1.discourse-cdn.com/flex002/uploads/python1/original/3X/5/6/560df2fd4fc9fc8090793583f6fefe2f74ae3c0a.png)

The previously mentioned working example with [`shutil.copy2()`](https://docs.python.org/3/library/shutil.html#shutil.copy2)

```python
import os, subprocess, shutil, sys, tempfile
os.makedirs(os.path.join(tempfile.gettempdir(), "modifield_python"), exist_ok=True)
renamed_pythonexefile = shutil.copy2(sys.executable, os.path.join(tempfile.gettempdir(), "modifield_python\\modified_python.exe"))

```

This script should produce `modified_python.exe` that upon double clicking it, will open python console.

Path where the produced file resides: `C:\Users\%USERNAME%\AppData\Local\Temp\modifield_python\modified_python.exe`

For me, it seems the issue is solved.

---

<div class="post-metadata">

**Author:** ![barneygale](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/barneygale/32/1882_2.png) [@barneygale](https://discuss.python.org/u/barneygale)\
**Post date:** [July 12, 2024, 7:40pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/6 "2024-07-12T19:40:04Z")

</div>

`copyfile()` calls `_copyfileobj_readinto()` on Windows:

> <https://github.com/python/cpython/blob/dc03ce797ae8786a9711e6ee5dcaadde02c55864/Lib/shutil.py#L176-L194>

Is it possible that `readinto()` can read `n < length` bytes _without_ reaching the end of the file? If so I think the `break` on line 192 might be at fault.

---

<div class="post-metadata">

**Author:** ![MegaIng](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/megaing/32/16162_2.png) [@MegaIng](https://discuss.python.org/u/MegaIng)\
**Post date:** [July 12, 2024, 8:32pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/7 "2024-07-12T20:32:25Z")

</div>

This ends up at a single call to [`_Py_read`](https://github.com/python/cpython/blob/dc03ce797ae8786a9711e6ee5dcaadde02c55864/Python/fileutils.c#L1859) from what I can tell, for which the docstring only states that EOF can be assumed when the return value is 0.

> ```python
> On success, return the number of read bytes, it can be lower than count.
> If the current file offset is at or past the end of file, no bytes are read,
> and read() returns zero.
> 
> ```

So yes, from what I can tell it’s invalid to assume that `readinto` returning less than `length` bytes means we are at an EOF since C doesn’t guarantee that behavior for the builtin `read`, and python doesn’t do anything to guarantee that behavior instead.

---

<div class="post-metadata">

**Author:** ![barry-scott](https://avatars.discourse-cdn.com/v4/letter/b/e9c0ed/32.png) [@barry-scott](https://discuss.python.org/u/barry-scott)\
**Post date:** [July 12, 2024, 9:14pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/8 "2024-07-12T21:14:51Z")

</div>

According to your info.  
You are comparing python.exe to a copy of pythonw.exe they will not be the same.

---

<div class="post-metadata">

**Author:** ![BoQsc](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/boqsc/32/20296_2.png) [@BoQsc](https://discuss.python.org/u/BoQsc)\
**Post date:** [July 13, 2024, 4:48am UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/9 "2024-07-13T04:48:00Z")

</div>

Well, it seems like the mystery is solved. I’ve compared against pythonw, and yeah, they are identical. This is overlook on my part as I’ve overfocused on the metadata, not realizing that I’m using statement `sys.executable` which becomes either `pythonw.exe` or `python.exe` and they are different in size.

---

<div class="post-metadata">

**Author:** ![eryksun](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/eryksun/32/697_2.png) [@eryksun](https://discuss.python.org/u/eryksun)\
**Post date:** [July 13, 2024, 6:38am UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/10 "2024-07-13T06:38:38Z")

</div>

Note that executing “modified\_python.exe”, as copied by your example, requires Python’s installation directory to be in the `PATH` environment variable, in order to find the dependent DLLs “python312.dll” and “vcruntime140.dll”. Normally this is not required because they’re located in the directory that contains the executable – “python.exe” or “pythonw.exe”, which the system searches implicitly. To avoid the extra `PATH` requirement, you’ll have to symlink, hardlink, or copy those two DLLs to the temp directory that contains “modified\_python.exe”.

---

<div class="post-metadata">

**Author:** ![BoQsc](https://sea2.discourse-cdn.com/flex002/user_avatar/discuss.python.org/boqsc/32/20296_2.png) [@BoQsc](https://discuss.python.org/u/BoQsc)\
**Post date:** [July 24, 2024, 2:54pm UTC](https://discuss.python.org/t/shutil-copyfile-corrupts-copied-file-when-running-under-pythonw-on-windows-10-home/58019/11 "2024-07-24T14:54:04Z")

</div>

Here is the utiilty script.

```python
usage: windows_create_renamed_copy_of_python_executable.py [-h] file_path

Produce renamed Python Executable on Windows Operating System.

positional arguments:
  file_path Path to the new executable file

options:
  -h, --help show this help message and exit

Example usage: python windows_create_renamed_copy_of_python_executable.py ".\test.exe"

```

#### `windows_create_renamed_copy_of_python_executable.py`

```python
def produce_renamed_python_executable(new_executable_path: str):
    import os, shutil, pathlib, sys
    python_executable_installed = pathlib.Path(sys.exec_prefix, 'python.exe') # change to pythonw to create a pythonw executable.
    python_executable_new = pathlib.Path(new_executable_path)
    os.makedirs(os.path.dirname(new_executable_path), exist_ok=True)
    return str(os.path.abspath(shutil.copy2(python_executable_installed, python_executable_new)))

def main():
    import argparse
    parser = argparse.ArgumentParser(
        description='Produce renamed Python Executable on Windows Operating System.', 
        epilog='Example usage: python %(prog)s ".\\test.exe"')
    parser.add_argument('file_path', type=str, help='Path to the new executable file')
    try:
        new_executable_path = produce_renamed_python_executable(parser.parse_args().file_path)
        print(f"Successfully created new executable at: {new_executable_path}")
        return 0
    except FileNotFoundError:
        print("Error: Source Python executable not found.", file=sys.stderr)
        return 1
    except PermissionError:
        print("Error: Permission denied. Make sure you have the necessary rights.", file=sys.stderr)
        return 2
    except Exception as e:
        print(f"Error: An unexpected error occurred: {str(e)}", file=sys.stderr)
        return 3

if __name__ ==" __main__": 
    import sys
    sys.exit(main())

```