avoid pycuda import warnings on win32

packaging/MSWindows/MINGW_BUILD.sh

@@ -502,6 +502,11 @@ if [ "${BUNDLE_DESKTOPLOGON}" == "1" ]; then
 	done
 fi
 
+if [ "${DO_CUDA}" == "1" ]; then
+	#pycuda wants a CUDA_PATH with "/bin" in it:
+	mkdir "${DIST}/bin"
+fi
+
 if [ "${DO_VERPATCH}" == "1" ]; then
 	for exe in `ls dist/*exe | grep -v Plink.exe`; do
 		tool_name=`echo $exe | sed 's+dist/++g;s+Xpra_++g;s+Xpra-++g;s+_+ +g;s+-+ +g;s+\.exe++g'`

xpra/codecs/cuda_common/cuda_context.py

@@ -16,11 +16,14 @@
 from xpra.util import engs, print_nested_dict, envint, csv, first_time
 from xpra.platform.paths import (
     get_default_conf_dirs, get_system_conf_dirs, get_user_conf_dirs,
-    get_resources_dir,
+    get_resources_dir, get_app_dir,
     )
-from xpra.os_util import load_binary_file
+from xpra.os_util import load_binary_file, WIN32
 from xpra.log import Logger
 
+if WIN32 and not os.environ.get("CUDA_PATH"):
+    os.environ["CUDA_PATH"] = os.path.join(get_app_dir(), "bin")
+
 with numpy_import_lock:
     import pycuda               #@UnresolvedImport
     from pycuda import driver   #@UnresolvedImport