What causes "CreateTexture2D returned no texture" in screenpipe/screenpipe?

Update/clean-reinstall GPU drivers

screenpipe/screenpipe · error · anyhow::Error

CreateTexture2D returned no texture

Error message

CreateTexture2D returned no texture

What it means

This is the companion null-check for [553]: `CreateTexture2D` returned S_OK but left the output texture pointer null. Since downstream code (CopyResource, Map) cannot tolerate a null texture, `create_texture_like` converts it into an explicit error. It is a defensive invariant check that rarely fires; when it does, it points to a device/runtime anomaly.

Source

Thrown at crates/screenpipe-screen/src/wgc_capture.rs:329

fn create_texture_like(
    device: &ID3D11Device,
    src_desc: &D3D11_TEXTURE2D_DESC,
    usage: D3D11_USAGE,
    cpu_access_flags: u32,
) -> Result<ID3D11Texture2D> {
    let mut desc = *src_desc;
    desc.Usage = usage;
    desc.BindFlags = 0;
    desc.CPUAccessFlags = cpu_access_flags;
    desc.MiscFlags = 0;

    let mut texture = None;
    unsafe {
        device
            .CreateTexture2D(&desc, None, Some(&mut texture))
            .map_err(|e| anyhow!("CreateTexture2D: {}", e))?;
    }
    texture.ok_or_else(|| anyhow!("CreateTexture2D returned no texture"))
}

extern "system" fn monitor_enum_proc(
    h_monitor: HMONITOR,
    _hdc: HDC,
    _rect: *mut RECT,
    state: LPARAM,
) -> BOOL {
    unsafe {
        (*(state.0 as *mut Vec<HMONITOR>)).push(h_monitor);
    }
    TRUE
}

/// Resolve a live `HMONITOR` for `monitor_id` (xcap's monitor id is `HMONITOR.0 as u32`).
/// Re-enumerating (rather than reconstructing the pointer from the id) makes sure the
/// handle is still valid right now, matching how the rest of this crate resolves ids.
fn find_hmonitor(monitor_id: u32) -> Result<HMONITOR> {

View on GitHub (pinned to 4ebf712990)

Solutions

Update/clean-reinstall GPU drivers
Disable overlay/recording software that may hook d3d11.dll and retry
Run `sfc /scannow` to verify d3d11.dll integrity
Log the descriptor (size/format) that triggered it to rule out a degenerate request
Treat as fatal for the capture session and restart the WGC capture

Defensive patterns

Strategy: try-catch

Try / catch

match create_texture_like(&device, &src_desc, staging) {
    Err(e) if e.to_string().contains("returned no texture") => {
        // environment anomaly: restart session, flag d3d11.dll hooking
        eprintln!("CreateTexture2D silently failed; restarting capture session");
        restart_capture_session();
    }
    Ok(t) => { /* use texture */ }
    Err(e) => return Err(e),
}

Prevention

Treat silent nulls as a session-level fault and restart the WGC capture
Check for DLL hooks on d3d11.dll in affected environments
Keep GPU drivers clean and current
Log the descriptor alongside the failure for reproducibility

When it happens

Trigger: `device.CreateTexture2D(&desc, None, Some(&mut texture))` reports success but `texture` stays `None`, reached from `store_frame` or `readback` — e.g. under DLL hooking of d3d11.dll or a broken driver that fakes success.

Common situations: Same environments as [553]: hooked d3d11.dll (overlays/AV), corrupt driver installations, or exotic third-party graphics-stack shims that return success without creating the object.

Related errors

AI-assisted analysis of screenpipe/screenpipe@4ebf712990 (2026-09-01). Data as JSON: /api/errors/94ee012462f5dde1. Report an issue: GitHub.