Picker crashes immediately on 5.6.0 in RDP sessions (0xc0000005, regression from 5.5.5)

[mukkino]

• bt.exe 5.6.0 picker crashes with exception 0xc0000005 at fault offset 0x0000000000094243
• Every single picker invocation crashes at the same offset
• Regression from 5.5.5 which works correctly
• Running via RDP session with AMD Radeon graphics + Microsoft Remote Display Adapter

[mukkino]

1. Reproduction steps

1. Install BT 5.6.0 via winget on a session using Microsoft Remote Display Adapter (RDP) + AMD Radeon
2. Set always = y in [picker]
3. Click any link
4. Picker does not appear

2. Crash evidence from Windows Event Log

• Faulting application: bt.exe, version 5.6.0.0
• Faulting module: bt.exe (self — not a DLL)
• Exception code: 0xc0000005 (access violation / null dereference)
• Fault offset: 0x0000000000094243

This offset is identical on every single crash — 100% deterministic, not flaky.

3. What the log tells us

[timestamp] showing rule picker for [url]
[timestamp] started with args: [url]|[hwnd]

The picker process starts, logs one line, then crashes. No window ever appears. Our Win32 EnumWindows scan confirmed no picker window exists at any point.

4. Root cause hypothesis (with source code reference)

From picker_app.cpp line 20:

app{grey::app::make(title, 100, 100)},

Immediately followed by:

app->win32_can_resize = false;   // crash here if make() returned null
app->win32_center_on_screen = true;

If grey::app::make() returns null (or an empty unique_ptr) on this display configuration, the first dereference produces a deterministic access violation. The config_app (main BT window) works fine — same grey framework — so something changed specifically in how make() initialises for the picker context in 5.6.0.

5. Confirmation it's a regression

• BT 5.5.5 works perfectly on the same machine, same session, same display adapter
• Immediate rollback to 5.5.5 via winget install --version 5.5.5 resolves it

6. Display/GPU config (already ruled out as root cause)

• Microsoft Remote Display Adapter (RDP session) + AMD Radeon 31.0.21924.61 (Dec 2025 driver)
• Parsec Virtual Display Driver also tested — same crash at same offset

Therefore it's not a display adapter issue

7. What makes no difference

• show_native_chrome = y or n — same crash either way
• close_on_focus_loss = n — same crash
• Icon paths in config.ini — same crash

[aloneguid]

whcih windows version are you using?

[mukkino]

OsName = Microsoft Windows 11 Pro
OSDisplayVersion = 25H2
OsBuildNumber = 26200
UBR = 8037
Windows 11 Pro 25H2, build 26200.8037

[aloneguid]

I can't reproduce this in Windows sandbox but will keep looking.

[mukkino]

Thank you

[aloneguid]

Do you mind trying 5.6.2 @mukkino ?