Need help

[snolan123]

I seriously cannot figure this out. Here is the server code:

rfbScreenInfoPtr server = rfbGetScreen(nullptr, nullptr, 8, 8, 8, 3, 4);
if (!server)
				return 1;
server->frameBuffer = (char*)malloc(8 * 8 * 4);
std::fill_n(server->frameBuffer, 8 * 8 * 4, 0xff);

print_framebuffer_bits((uint8_t*)server->frameBuffer, 8, 8);

rfbInitServer(server);
vnc_thread = std::thread(vnc_server_thread_func, server);`

And here is the resize function for the client

static rfbBool resize(rfbClient* client) {
    int width = client->width;
    int height = client->height;
    int depth = client->format.bitsPerPixel;

    // 1. Clean up old buffer allocations
    if (vncBuffer) {
        delete[] vncBuffer;
        vncBuffer = nullptr;
    }

    // 2. Allocate a tightly packed buffer (Width * 4 bytes, ZERO alignment padding)
    vncBuffer = new uint8_t[width * height * 4];
    std::memset(vncBuffer, 0, width * height * 4);

    rfbClientSetClientData(client, SDL_Init, vncBuffer);

    // 3. Bind LibVNC straight to this raw buffer
    client->frameBuffer = vncBuffer;

    SetFormatAndEncodings(client);

    // 5. Recreate the streaming texture using XRGB8888
    // This aligns the little-endian bytes perfectly to [Pad][B][G][R]
    if (sdlTexture)
        SDL_DestroyTexture(sdlTexture);

    sdlTexture = SDL_CreateTexture(renderer,
        SDL_PIXELFORMAT_XRGB8888,
        SDL_TEXTUREACCESS_STREAMING,
        width, height);

    if (!sdlTexture)
        std::printf("resize: error creating texture: %s\n", SDL_GetError());

    return TRUE;
}

No matter what I try the first two pixels on the first row are always black. Can someone tell me what I am doing wrong?

[bk138]

Please have a look at the examples, they are known to work.

[snolan123]

I'm looking at the examples but they are outdated. Are there plans to use SDL3?

[marcofortina]

This is not a row alignment or SDL pitch issue. Those two black pixels are the LibVNCServer default cursor being drawn into your framebuffer.

LibVNCServer installs a default cursor in rfbGetScreen(). That cursor is 8x7 with hotspot (3, 3). The initial cursor position is (0, 0), so the cursor is drawn starting at (-3, -3) and then clipped against your tiny 8x8 framebuffer.

That means the first framebuffer row receives row 3 of the default cursor, starting from cursor column 3. In that row, the first two visible cursor pixels are foreground pixels, and the default cursor foreground color is black. So the first two pixels on the first row become black even though your framebuffer was originally filled with 0xff.

Disable the default server-side cursor before calling rfbInitServer():

rfbScreenInfoPtr server = rfbGetScreen(nullptr, nullptr, 8, 8, 8, 3, 4);
if (!server)
    return 1;

server->cursor = nullptr;   // Disable the default server-side cursor

server->frameBuffer = static_cast<char*>(malloc(8 * 8 * 4));
if (!server->frameBuffer)
    return 1;

std::fill_n(server->frameBuffer, 8 * 8 * 4, 0xff);

print_framebuffer_bits(reinterpret_cast<uint8_t*>(server->frameBuffer), 8, 8);

rfbInitServer(server);
vnc_thread = std::thread(vnc_server_thread_func, server);

Alternatively, make the client request remote cursor updates instead of letting the server draw the cursor into the framebuffer:

client->appData.useRemoteCursor = TRUE;

Make sure this is set before SetFormatAndEncodings(client) is sent.

Your resize function can stay mostly as-is, but I would also make the pixel format explicit if you are going to upload the buffer to an SDL_PIXELFORMAT_XRGB8888 texture:

static rfbBool resize(rfbClient* client)
{
    const int width = client->width;
    const int height = client->height;

    delete[] vncBuffer;
    vncBuffer = nullptr;

    vncBuffer = new uint8_t[width * height * 4];
    std::memset(vncBuffer, 0, width * height * 4);

    client->frameBuffer = vncBuffer;

    client->format.bitsPerPixel = 32;
    client->format.depth = 24;
    client->format.trueColour = TRUE;

    client->format.redMax = 255;
    client->format.greenMax = 255;
    client->format.blueMax = 255;

    client->format.redShift = 16;
    client->format.greenShift = 8;
    client->format.blueShift = 0;

    client->appData.useRemoteCursor = TRUE;

    SetFormatAndEncodings(client);

    if (sdlTexture)
        SDL_DestroyTexture(sdlTexture);

    sdlTexture = SDL_CreateTexture(renderer,
        SDL_PIXELFORMAT_XRGB8888,
        SDL_TEXTUREACCESS_STREAMING,
        width,
        height);

    if (!sdlTexture)
        std::printf("resize: error creating texture: %s\n", SDL_GetError());

    return TRUE;
}

And when updating the SDL texture, use the correct pitch:

SDL_UpdateTexture(sdlTexture, nullptr, vncBuffer, client->width * 4);

The important fix for the two black pixels is the server-side one:

server->cursor = nullptr;

If those pixels disappear after adding that line, then the framebuffer was fine and what you were seeing was simply the default cursor clipped into the top-left corner.

[snolan123]

Thank you. After much testing I finally came to that same conclusion and manage to disable the cursor. Those are not well documented though

…

On Thu, 11 Jun 2026 at 00:37, Marco Fortina ***@***.***> wrote: *marcofortina* left a comment (LibVNC/libvncserver#726) <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_LibVNC_libvncserver_issues_726-23issuecomment-2D4670254403&d=DwMFaQ&c=_Zm2Kb9yJrso4r0-N78Ezl-a6DfgMbIISn5tZXXqWk4&r=4Ga46s1ZOJAzHgMvGZD27UiTRFz6bNOaOejvQZvmz98&m=CEM4oR4Ekhb14VX3ezD8T8KCxIUDHSVoWa8BDpCS0mw-sWToaxqr9wrXR_KTdW4k&s=4FU8AZ5Lc1slGjgXmL8rOiiHTsTz6HfcBONps7HAIws&e=> This is not a row alignment or SDL pitch issue. Those two black pixels are the LibVNCServer default cursor being drawn into your framebuffer. LibVNCServer installs a default cursor in rfbGetScreen(). That cursor is 8x7 with hotspot (3, 3). The initial cursor position is (0, 0), so the cursor is drawn starting at (-3, -3) and then clipped against your tiny 8x8 framebuffer. That means the first framebuffer row receives row 3 of the default cursor, starting from cursor column 3. In that row, the first two visible cursor pixels are foreground pixels, and the default cursor foreground color is black. So the first two pixels on the first row become black even though your framebuffer was originally filled with 0xff. Disable the default server-side cursor before calling rfbInitServer(): rfbScreenInfoPtr server = rfbGetScreen(nullptr, nullptr, 8, 8, 8, 3, 4);if (!server) return 1; server->cursor = nullptr; // Disable the default server-side cursor server->frameBuffer = static_cast<char*>(malloc(8 * 8 * 4));if (!server->frameBuffer) return 1; std::fill_n(server->frameBuffer, 8 * 8 * 4, 0xff); print_framebuffer_bits(reinterpret_cast<uint8_t*>(server->frameBuffer), 8, 8); rfbInitServer(server); vnc_thread = std::thread(vnc_server_thread_func, server); Alternatively, make the client request remote cursor updates instead of letting the server draw the cursor into the framebuffer: client->appData.useRemoteCursor = TRUE; Make sure this is set before SetFormatAndEncodings(client) is sent. Your resize function can stay mostly as-is, but I would also make the pixel format explicit if you are going to upload the buffer to an SDL_PIXELFORMAT_XRGB8888 texture: static rfbBool resize(rfbClient* client) { const int width = client->width; const int height = client->height; delete[] vncBuffer; vncBuffer = nullptr; vncBuffer = new uint8_t[width * height * 4]; std::memset(vncBuffer, 0, width * height * 4); client->frameBuffer = vncBuffer; client->format.bitsPerPixel = 32; client->format.depth = 24; client->format.trueColour = TRUE; client->format.redMax = 255; client->format.greenMax = 255; client->format.blueMax = 255; client->format.redShift = 16; client->format.greenShift = 8; client->format.blueShift = 0; client->appData.useRemoteCursor = TRUE; SetFormatAndEncodings(client); if (sdlTexture) SDL_DestroyTexture(sdlTexture); sdlTexture = SDL_CreateTexture(renderer, SDL_PIXELFORMAT_XRGB8888, SDL_TEXTUREACCESS_STREAMING, width, height); if (!sdlTexture) std::printf("resize: error creating texture: %s\n", SDL_GetError()); return TRUE; } And when updating the SDL texture, use the correct pitch: SDL_UpdateTexture(sdlTexture, nullptr, vncBuffer, client->width * 4); The important fix for the two black pixels is the server-side one: server->cursor = nullptr; If those pixels disappear after adding that line, then the framebuffer was fine and what you were seeing was simply the default cursor clipped into the top-left corner. — Reply to this email directly, view it on GitHub <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_LibVNC_libvncserver_issues_726-3Femail-5Fsource-3Dnotifications-26email-5Ftoken-3DBKK5XFGQYGZ3Y324SDUKCC347FJBLA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINRXGAZDKNBUGAZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L-23issuecomment-2D4670254403&d=DwMFaQ&c=_Zm2Kb9yJrso4r0-N78Ezl-a6DfgMbIISn5tZXXqWk4&r=4Ga46s1ZOJAzHgMvGZD27UiTRFz6bNOaOejvQZvmz98&m=CEM4oR4Ekhb14VX3ezD8T8KCxIUDHSVoWa8BDpCS0mw-sWToaxqr9wrXR_KTdW4k&s=K1gQ94ASBhn3nS-2xx0gUTofqMkFo7EH-MCAspvS51M&e=>, or unsubscribe <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_BKK5XFDL7N3PYNDEICKBDHT47FJBLAVCNFSNUABEKJSXA33TNF2G64TZHMYTQMRXG44DQNZ3JFZXG5LFHM2DMMJYGMYTAMRYGCQXMAQ&d=DwMFaQ&c=_Zm2Kb9yJrso4r0-N78Ezl-a6DfgMbIISn5tZXXqWk4&r=4Ga46s1ZOJAzHgMvGZD27UiTRFz6bNOaOejvQZvmz98&m=CEM4oR4Ekhb14VX3ezD8T8KCxIUDHSVoWa8BDpCS0mw-sWToaxqr9wrXR_KTdW4k&s=2h_IVsQwFtCWrEmfeF0DfFElCjEhmCR9IJwdJDUpddQ&e=> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_mobile_ios_BKK5XFGH5F6GBRHUYFGMGED47FJBLA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINRXGAZDKNBUGAZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG&d=DwMFaQ&c=_Zm2Kb9yJrso4r0-N78Ezl-a6DfgMbIISn5tZXXqWk4&r=4Ga46s1ZOJAzHgMvGZD27UiTRFz6bNOaOejvQZvmz98&m=CEM4oR4Ekhb14VX3ezD8T8KCxIUDHSVoWa8BDpCS0mw-sWToaxqr9wrXR_KTdW4k&s=fwgEMNuLL-2Cq4szU2WfkGUF7uCHWwQwFFdpsx1_pBY&e=> and Android <https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_mobile_android_BKK5XFA45UKGLGZ4I2UMOML47FJBLA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINRXGAZDKNBUGAZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA&d=DwMFaQ&c=_Zm2Kb9yJrso4r0-N78Ezl-a6DfgMbIISn5tZXXqWk4&r=4Ga46s1ZOJAzHgMvGZD27UiTRFz6bNOaOejvQZvmz98&m=CEM4oR4Ekhb14VX3ezD8T8KCxIUDHSVoWa8BDpCS0mw-sWToaxqr9wrXR_KTdW4k&s=R9qUlV66Hcao8IQI6Gwa5R4AfX56Jehmx5Z18Tj-ZAE&e=>. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

[snolan123]

Another question. I can see the server desktop and it updates continuously but I can't see the mouse pointer. How can I see the server mouse pointer?

[marcofortina]

Yes, that is expected depending on the cursor mode you selected.

In VNC there are two different cursor modes:

1. The server draws the cursor directly into the framebuffer.
2. The server sends cursor shape/position updates, and the viewer draws the cursor itself.

If you used this line from the previous test:

server->cursor = nullptr;

then you removed the server cursor completely, so there is no cursor to draw or send.

If you want the simplest behaviour, remove that line and do not request remote cursor support on the client:

client->appData.useRemoteCursor = FALSE;
SetFormatAndEncodings(client);

With useRemoteCursor = FALSE, LibVNCServer will draw the cursor into the framebuffer for that client.

If instead you set:

client->appData.useRemoteCursor = TRUE;

then you are telling the server: “send me cursor shape and cursor position updates; do not draw the cursor into the framebuffer.” In that mode your SDL viewer must draw the cursor as an overlay.

A minimal client-side setup would look like this:

static int cursorX = 0;
static int cursorY = 0;
static int cursorHotX = 0;
static int cursorHotY = 0;
static int cursorW = 0;
static int cursorH = 0;
static int cursorBpp = 0;

static std::vector<uint8_t> cursorPixels;
static std::vector<uint8_t> cursorMask;

static void got_cursor_shape(rfbClient* client,
                             int xhot,
                             int yhot,
                             int width,
                             int height,
                             int bytesPerPixel)
{
    cursorHotX = xhot;
    cursorHotY = yhot;
    cursorW = width;
    cursorH = height;
    cursorBpp = bytesPerPixel;

    cursorPixels.assign(client->rcSource,
                        client->rcSource + width * height * bytesPerPixel);

    cursorMask.assign(client->rcMask,
                      client->rcMask + width * height);
}

static rfbBool handle_cursor_pos(rfbClient* client, int x, int y)
{
    cursorX = x;
    cursorY = y;
    return TRUE;
}

Register the callbacks before sending SetFormatAndEncodings():

client->appData.useRemoteCursor = TRUE;
client->GotCursorShape = got_cursor_shape;
client->HandleCursorPos = handle_cursor_pos;

SetFormatAndEncodings(client);

Then, in your SDL rendering code, draw the desktop framebuffer first and draw the cursor afterwards at:

cursorX - cursorHotX,
cursorY - cursorHotY

using cursorPixels as the cursor image and cursorMask as the visibility mask.

So the quick answer is:

• If you want LibVNCServer to draw the pointer into the framebuffer, use client->appData.useRemoteCursor = FALSE and do not set server->cursor = nullptr.
• If you use client->appData.useRemoteCursor = TRUE, then the pointer will not appear automatically in your SDL texture. You must render it yourself using GotCursorShape and HandleCursorPos.

[snolan123]

I tried both ways client->appData.useRemoteCursor = FALSE didn't do anything and the other way just showed me the default X pointer