SDL fallback, partition fix, lz4 remote streaming, delay-oobe-run

SDL display:
- try SDL, auto-fallback to headless if it fails (no crash)
- SDL_VIDEODRIVER=x11 to avoid wayland socket path issues
- suppress XDG_RUNTIME_DIR warnings

Disk copy:
- zap-all before writing to clear old partition tables
- delete recovery partition (4) before resizing partition 3
- use parted resizepart (preserves partition GUID for BCD)
- remote: nix-shell for sgdisk/parted/ntfsresize on target
- remote: lz4 compression for faster streaming
- remote: pv progress bar with disk size
- -y/--yes flag to skip confirmation prompt

Generalize:
- delay-oobe-run=true defers OOBE + activation to real hardware
- clean cached Autounattend from Windows\Panther before sysprep
- taskkill sysprep.exe on first login (CopyProfile artifact)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

flake.nix

@@ -126,6 +126,7 @@
       GENERALIZE=""
       TO_DISK=""
       TO_REMOTE_DISK=""
+      YES=false
       OUT_LINK="./result"
 
       while [[ ''${#} -gt 0 ]]; do
@@ -134,6 +135,7 @@
           --generalize)      GENERALIZE="$2"; shift 2 ;;
           --to-disk)         TO_DISK="$2"; shift 2 ;;
           --to-remote-disk)  TO_REMOTE_DISK="$2"; shift 2 ;;
+          -y|--yes)          YES=true; shift ;;
           --out-link)        OUT_LINK="$2"; shift 2 ;;
           --help|-h)         usage ;;
           *)                 echo "Unknown option: $1"; usage ;;
@@ -212,17 +214,24 @@
         echo "Target: $TO_DISK ($DISK_SIZE_GB GB)"
         echo ""
         echo "WARNING: This will DESTROY all data on $TO_DISK"
-        read -rp "Continue? [y/N] " confirm
-        [[ "$confirm" != "y" && "$confirm" != "Y" ]] && { echo "Aborted."; exit 0; }
+        if [[ "$YES" != "true" ]]; then
+          read -rp "Continue? [y/N] " confirm
+          [[ "$confirm" != "y" && "$confirm" != "Y" ]] && { echo "Aborted."; exit 0; }
+        fi
 
         echo ""
-        echo "[1/4] Writing image to disk..."
+        echo "[1/5] Wiping partition table..."
+        ${pkgs.gptfdisk}/bin/sgdisk --zap-all "$TO_DISK"
+
+        echo "[2/5] Writing image to disk..."
         ${pkgs.qemu}/bin/qemu-img convert -p -S 4k -f qcow2 -O raw "$IMAGE_FILE" "$TO_DISK"
 
-        echo "[2/4] Fixing GPT backup header..."
+        echo "[3/5] Fixing GPT backup header..."
         ${pkgs.gptfdisk}/bin/sgdisk -e "$TO_DISK"
+        # delete recovery partition if present, then resize Windows partition
+        ${pkgs.gptfdisk}/bin/sgdisk -d 4 "$TO_DISK" 2>/dev/null || true
 
-        echo "[3/4] Expanding Windows partition (partition 3) to fill disk..."
+        echo "[4/5] Expanding Windows partition (partition 3)..."
         ${pkgs.parted}/bin/parted -s "$TO_DISK" resizepart 3 100%
 
         if [[ "$TO_DISK" == *nvme* ]] || [[ "$TO_DISK" == *mmcblk* ]]; then
@@ -231,7 +240,7 @@
           WIN_PART="''${TO_DISK}3"
         fi
 
-        echo "[4/4] Expanding NTFS filesystem on $WIN_PART..."
+        echo "[5/5] Expanding NTFS filesystem on $WIN_PART..."
         ${pkgs.ntfs3g}/bin/ntfsresize --force --no-action "$WIN_PART"
         echo "y" | ${pkgs.ntfs3g}/bin/ntfsresize --force "$WIN_PART"
 
@@ -252,19 +261,28 @@
         echo "Disk:   $REMOTE_DISK"
         echo ""
         echo "WARNING: This will DESTROY all data on $REMOTE_HOST:$REMOTE_DISK"
-        read -rp "Continue? [y/N] " confirm
-        [[ "$confirm" != "y" && "$confirm" != "Y" ]] && { echo "Aborted."; exit 0; }
+        if [[ "$YES" != "true" ]]; then
+          read -rp "Continue? [y/N] " confirm
+          [[ "$confirm" != "y" && "$confirm" != "Y" ]] && { echo "Aborted."; exit 0; }
+        fi
 
         echo ""
-        echo "[1/4] Streaming image to remote disk..."
-        ${pkgs.qemu}/bin/qemu-img convert -f qcow2 -O raw "$IMAGE_FILE" /dev/stdout \
-          | ssh "$REMOTE_HOST" "dd of=$REMOTE_DISK bs=4M status=progress oflag=direct"
+        echo "[1/5] Wiping partition table..."
+        ssh "$REMOTE_HOST" "sgdisk --zap-all $REMOTE_DISK"
 
-        echo "[2/4] Fixing GPT backup header..."
-        ssh "$REMOTE_HOST" "sgdisk -e $REMOTE_DISK"
-
-        echo "[3/4] Expanding Windows partition (partition 3) to fill disk..."
-        ssh "$REMOTE_HOST" "parted -s $REMOTE_DISK resizepart 3 100%"
+        echo "[2/5] Streaming image to remote disk..."
+        NBD_SOCK="/tmp/vmix-nbd-$$.sock"
+        ${pkgs.qemu}/bin/qemu-nbd --read-only -f qcow2 -k "$NBD_SOCK" "$IMAGE_FILE" &
+        NBD_PID=$!
+        trap "kill $NBD_PID 2>/dev/null; rm -f $NBD_SOCK" EXIT
+        while [ ! -S "$NBD_SOCK" ]; do sleep 0.1; done
+        DISK_SIZE=$(${pkgs.libnbd}/bin/nbdinfo --size "nbd+unix:///?socket=$NBD_SOCK")
+        ${pkgs.libnbd}/bin/nbdcopy --request-size=4194304 "nbd+unix:///?socket=$NBD_SOCK" - \
+          | ${pkgs.pv}/bin/pv -s "$DISK_SIZE" \
+          | ${pkgs.lib.getBin pkgs.lz4}/bin/lz4 -1 - \
+          | ssh "$REMOTE_HOST" "nix-shell -p lz4 --run 'lz4 -d - - | dd of=$REMOTE_DISK bs=4M iflag=fullblock oflag=direct conv=sparse'"
+        kill $NBD_PID 2>/dev/null || true
+        rm -f "$NBD_SOCK"
 
         if [[ "$REMOTE_DISK" == *nvme* ]] || [[ "$REMOTE_DISK" == *mmcblk* ]]; then
           REMOTE_WIN_PART="''${REMOTE_DISK}p3"
@@ -272,8 +290,14 @@
           REMOTE_WIN_PART="''${REMOTE_DISK}3"
         fi
 
-        echo "[4/4] Expanding NTFS filesystem on $REMOTE_WIN_PART..."
-        ssh "$REMOTE_HOST" "echo y | ntfsresize --force $REMOTE_WIN_PART"
+        echo "[3/5] Fixing GPT backup header..."
+        ssh "$REMOTE_HOST" "nix-shell -p gptfdisk --run 'sgdisk -e $REMOTE_DISK && sgdisk -d 4 $REMOTE_DISK 2>/dev/null || true'"
+
+        echo "[4/5] Expanding Windows partition (partition 3)..."
+        ssh "$REMOTE_HOST" "nix-shell -p parted --run 'parted -s $REMOTE_DISK resizepart 3 100%'"
+
+        echo "[5/5] Expanding NTFS filesystem on $REMOTE_WIN_PART..."
+        ssh "$REMOTE_HOST" "nix-shell -p ntfs3g --run 'echo y | ntfsresize --force $REMOTE_WIN_PART'"
 
         echo ""
         echo "Done. $REMOTE_HOST:$REMOTE_DISK is ready to boot."

lib/images/windows/helpers/customizeImage.nix

@@ -80,32 +80,36 @@
       chmod +w vars.fd
 
       VMIX_DISPLAY="-nographic"
-      ${lib.optionalString (displayArg != null) ''VMIX_DISPLAY="${displayArg}"''}
-      ${lib.optionalString (displayArg == null) ''
+      ${lib.optionalString (vncDisplay != null) ''VMIX_DISPLAY="-vnc ${vncDisplay}"''}
+      ${lib.optionalString (vncDisplay == null) ''
         VMIX_DF=$(ls -t /tmp/.vmix-display-* 2>/dev/null | head -1)
         if [ -n "$VMIX_DF" ]; then
           export DISPLAY=$(cat "$VMIX_DF")
           export HOME=$(mktemp -d)
           export XDG_RUNTIME_DIR=$HOME
+          export SDL_VIDEODRIVER=x11
           VMIX_DISPLAY="-display sdl"
         fi
       ''}
 
       echo "=== vmix: booting Audit Mode for ${name} ==="
-      timeout 1800 qemu-system-x86_64 \
-        $VMIX_DISPLAY \
-        -accel kvm \
-        -m ${toString memSize} \
-        -smp ${toString smp} \
-        -cpu host \
-        -machine type=q35 \
+      QEMU_ARGS="-accel kvm -m ${toString memSize} -smp ${toString smp} -cpu host -machine type=q35 \
         -drive if=pflash,format=raw,readonly=on,file=${pkgs.OVMF.fd}/FV/OVMF_CODE.fd \
         -drive if=pflash,format=raw,file=vars.fd \
         -rtc base=localtime,clock=host \
         -device qemu-xhci -device usb-tablet \
         -drive file=${resultImg},format=qcow2,if=virtio \
         ${cdromArgs} \
-        -nic user,model=virtio-net-pci
+        -nic user,model=virtio-net-pci"
+
+      timeout 1800 qemu-system-x86_64 $VMIX_DISPLAY $QEMU_ARGS || \
+        if [[ "$VMIX_DISPLAY" == "-display sdl" ]]; then
+          echo "=== vmix: SDL failed, retrying headless ==="
+          cp ${pkgs.OVMF.fd}/FV/OVMF_VARS.fd vars.fd && chmod +w vars.fd
+          timeout 1800 qemu-system-x86_64 -nographic $QEMU_ARGS
+        else
+          exit 1
+        fi
 
       echo "=== vmix: audit script ${name} complete ==="
     '';

lib/images/windows/helpers/makeImage.nix

@@ -53,19 +53,14 @@ let
         export DISPLAY=$(cat "$VMIX_DF")
         export HOME=$(mktemp -d)
         export XDG_RUNTIME_DIR=$HOME
+        export SDL_VIDEODRIVER=x11
         VMIX_DISPLAY="-display sdl"
       fi
     ''}
 
     # Windows installs unattended, reboots into Audit Mode,
     # deletes cached Autounattend, shuts down → QEMU exits.
-    timeout 3600 qemu-system-x86_64 \
-      $VMIX_DISPLAY \
-      -accel kvm \
-      -m ${toString memSize} \
-      -smp ${toString smp} \
-      -cpu host \
-      -machine type=q35 \
+    QEMU_ARGS="-accel kvm -m ${toString memSize} -smp ${toString smp} -cpu host -machine type=q35 \
       -drive if=pflash,format=raw,readonly=on,file=${pkgs.OVMF.fd}/FV/OVMF_CODE.fd \
       -drive if=pflash,format=raw,file=vars.fd \
       -rtc base=localtime,clock=host \
@@ -73,7 +68,16 @@ let
       -drive file=disk.qcow2,format=qcow2,if=virtio \
       -drive file=${iso},media=cdrom,readonly=on \
       -drive file=${drivers.virtio-iso},media=cdrom,readonly=on \
-      -nic user,model=virtio-net-pci
+      -nic user,model=virtio-net-pci"
+
+    timeout 3600 qemu-system-x86_64 $VMIX_DISPLAY $QEMU_ARGS || \
+      if [[ "$VMIX_DISPLAY" == "-display sdl" ]]; then
+        echo "=== vmix: SDL failed, retrying headless ==="
+        cp ${pkgs.OVMF.fd}/FV/OVMF_VARS.fd vars.fd && chmod +w vars.fd
+        timeout 3600 qemu-system-x86_64 -nographic $QEMU_ARGS
+      else
+        exit 1
+      fi
 
     echo "=== vmix: ${name} install complete ==="
     mv disk.qcow2 $out

lib/images/windows/templates/generalize.nix

@@ -53,6 +53,12 @@ in
     reg add "HKCU\Control Panel\Desktop" /v WallpaperStyle /t REG_SZ /d "0" /f
     ''}
 
+    :: Kill sysprep if it was triggered via CopyProfile'd startup entries
+    taskkill /f /im sysprep.exe 2>nul
+    :: Clean any leftover RunOnce/Run entries from audit phase
+    reg delete "HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\RunOnce" /v "vmixAudit" /f 2>nul
+    reg delete "HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\Run" /v "vmixAudit" /f 2>nul
+
     :: Remove Edge AppxPackage for current user (runs in user context during OOBE)
     :: The app is already removed on one of the templates but a ghost appx entry remains that can only be deleted at the user level
     powershell -Command "Get-AppxPackage *MicrosoftEdge* | Remove-AppxPackage -ErrorAction SilentlyContinue"
@@ -148,7 +154,7 @@ in
     </unattend>
   '';
 in {
-  name = if delayOobeRun then "sealed" else "generalize";
+  name = if delayOobeRun then "generalize-delay-oobe" else "generalize";
   uploads = [
     { source = oobeXml; dest = "/oobe-unattend.xml"; }
     { source = postOobeScript; dest = "/post-oobe.cmd"; }
@@ -158,6 +164,10 @@ in {
   # generalize: sysprep + reboot into OOBE in the same QEMU session
   auditScript = ''
     @echo off
+    :: Remove cached Autounattend from initial install (contains Audit Mode reseal)
+    del /q C:\Windows\Panther\unattend.xml 2>nul
+    del /q C:\Windows\Panther\Unattend\unattend.xml 2>nul
+    del /q C:\Windows\System32\Sysprep\Panther\unattend.xml 2>nul
     C:\Windows\System32\Sysprep\sysprep.exe /generalize /oobe ${if delayOobeRun then "/shutdown" else "/reboot"} /quiet /unattend:C:\oobe-unattend.xml
   '';
 }