Re: [RFC uncompiled PATCH] cocoa: run qemu_init in the main thread

[Paolo Bonzini <pbonzini@redhat.com>]

On 3/7/22 16:34, Akihiko Odaki wrote:
> Thanks for this interesting suggestion. However I don't think this 
> improves the situation much. The main contribution of this change is 
> that elimination of display_init_sem but it is still necessary for 
> command line usage of the executable.
> 
> display_init_sem is kind of overloaded has two roles. One is to tell 
> that the QEMU is ready to initialize the display. The other is to tell 
> if it is going to initialize the display, which would not happen when it 
> is used entirely in the command line. The former role can be eliminated 
> by waiting for qemu_init, but the latter cannot be.

This is easy to account for.  On top of this patch:

diff --git a/ui/cocoa.m b/ui/cocoa.m
index 44d8ea7a39..3903fa4b9b 100644
--- a/ui/cocoa.m
+++ b/ui/cocoa.m
@@ -96,6 +96,7 @@ static DisplayChangeListener dcl = {
  static int last_buttons;
  static int cursor_hide = 1;
  static bool full_screen;
+static bool have_cocoa_ui;
  
  static bool stretch_video;
  static NSTextField *pauseLabel;
@@ -1899,6 +1900,11 @@ int main (int argc, char **argv) {
  
      /* Takes iothread lock, released in applicationDidFinishLaunching:.  */
      qemu_init(argc, argv);
+    if (!have_cocoa_ui) {
+         qemu_main_loop();
+         qemu_cleanup();
+         return 0;
+    }
  
      NSAutoreleasePool * pool = [[NSAutoreleasePool alloc] init];
  
@@ -2021,6 +2027,7 @@ static void cocoa_refresh(DisplayChangeListener *dcl)
  static void cocoa_display_init(DisplayState *ds, DisplayOptions *opts)
  {
      COCOA_DEBUG("qemu_cocoa: cocoa_display_init\n");
+    have_cocoa_ui = 1;
      if (opts->has_full_screen && opts->full_screen) {
          full_screen = 1;
      }

Paolo