Application crashes when file_dialog path encounters characters like æ ø å

[Jornexe]

Version of Dear PyGui

Version: 1.8.0 Operating System: Windows 10

My Issue/Question

Application crashes if you have a character like æ, ø, å in your path when you press OK or CANCEL.

To Reproduce

Steps to reproduce the behavior:

1. open a file_dialog
2. Navigate to a path with a character like ø, æ or å
3. Click on either ok or cancel
4. Application crashes

Expected behavior

Executes either callback or cancel_callback.

Standalone, minimal, complete and verifiable example

import dearpygui.dearpygui as dpg
import os

new_folder_name = "new_folder_with_øæå"
current_dir = os.getcwd()
new_folder_path = os.path.join(current_dir, new_folder_name)
if (os.path.exists(new_folder_path) == False):
    os.makedirs(new_folder_path)

dpg.create_context()

def callback(sender, app_data):
    print('OK was clicked.')

def cancel_callback(sender, app_data):
    print('Cancel was clicked.')
    print("Sender: ", sender)
    print("App Data: ", app_data)


width, height = 600,600
dpg.add_file_dialog(
    directory_selector=True, show=False, callback=callback, tag="dialog", 
    cancel_callback=cancel_callback, width=400 ,height=400 )


dpg.show_item("dialog")

dpg.create_viewport(title='File Name', width=width, height=height)

dpg.setup_dearpygui()
dpg.show_viewport()

dpg.start_dearpygui()
dpg.destroy_context()

[v-ein]

It crashes in mvFileDialog::getInfoDict(), in the very first call to PyDict_SetItemString(), which extracts file name from ImGuiFileDialog and feeds it into ToPyString() (and consequently PyUnicode_FromString()):

PyDict_SetItemString(dict, "file_path_name", mvPyObject(ToPyString(_instance.GetFilePathName())));

The problem is, GetFilePathName() returns an ANSI-encoded (single-byte) string, whereas PyUnicode_FromString() expects a UTF-8 strings and treats it as such. I believe it goes over the trailing zero or something like that. Or maybe it stumps upon an assertion of some kind and crashes.

I'm not sure whether ANSI-encoded strings are Windows-specific behavior or they show up on all platforms (I'd bet on Windows-specific). The bad news are, the file dialog doesn't even display such names correctly. That is, strings inside of file dialog must be UTF-8 rather than ANSI; it's a bug in ImGuiFileDialog and not in DPG. Not sure if we can fix it in ImGuiFileDialog though because ~it's a submodule repo~ - scratch that, even though it's a third-party library, it's committed directly into DPG. Technically we can fix it or even update ImGuiFileDialog from its source repo (if it works on old ImGui, that is).

[v-ein]

Another piece of wisdom can be found in this comment on #299.