1 files changed, 96 insertions, 0 deletions
diff --git a/results/classifier/118/all/1703795 b/results/classifier/118/all/1703795
new file mode 100644
index 000000000..3f56cb59e
--- /dev/null
+++ b/results/classifier/118/all/1703795
@@ -0,0 +1,96 @@
+graphic: 0.978
+register: 0.972
+user-level: 0.955
+peripherals: 0.949
+device: 0.946
+socket: 0.945
+architecture: 0.943
+debug: 0.943
+virtual: 0.942
+assembly: 0.940
+kernel: 0.939
+semantic: 0.937
+performance: 0.929
+files: 0.927
+boot: 0.927
+arm: 0.921
+vnc: 0.917
+hypervisor: 0.911
+risc-v: 0.911
+network: 0.905
+PID: 0.903
+TCG: 0.901
+permissions: 0.900
+mistranslation: 0.894
+KVM: 0.887
+ppc: 0.878
+VMM: 0.871
+x86: 0.833
+i386: 0.798
+
+Unable to release mouse in SDL2 mode
+
+Starting with commit 8f4ea9cd0b770dbe496d9d24f0ef8813fdbfe0d0 "sdl: prefer sdl2 over sdl1", I can no longer release mouse pointer grab unless I use --with-sdlabi=1.2 configure option.
+
+This easily reproduces in e.g. guest Kubuntu, when I let it start Xorg and then click into the QEMU window. After this the mouse is trapped and no matter how I combine Ctrl+Alt and motion of the cursor, the pointer never goes out from the window. When at the border, QEMU window switches from "Press Ctrl+Alt to exit grab" to "QEMU", i.e. it thinks that it has released the grab. But it hasn't really, so I have to go to VT1 and do "pkill qemu" from there to get my pointer back.
+
+Which version of SDL2 are you using? Could you please try to reproduce it with the latest version to see whether the problem has been fixed there?
+
+I'm using SDL2-2.0.5, which I believe is the latest already.
+
+I am seeing this on my system as well - the exact same symptoms.  Has anyone investigated this problem?
+
+I'm also seeing this problem with -vga vmware in case that matters.
+
+Additionally 2ec78706d188df7d3dab43d07b19b05ef7800a44 also broke keyboard with SDL1 so that e.g. in a Windows password prompt backspace and enter just inserts chars instead of doing actions so now the workaround to avoid this bug (--with-sdlabi=1.2) is also unusable.
+
+Can you file a separate bug for the SDL1  backspace problem - it was not intended to cause problems like that.
+
+On Thu, 1 Feb 2018, Daniel Berrange wrote:
+> Can you file a separate bug for the SDL1  backspace problem - it was not
+> intended to cause problems like that.
+
+Unless you want it tracked in a bug I'm happy with a fix submitted to the 
+list eventually without being separately tracked on launchpad. I would've 
+reported this on the mailing list normally but since this is also relevant 
+here due to SDL2 also not uasable beacause of this bug I've commented 
+here so people affected by this know about both problems.
+
+
+workaround: don't click into the guest window before the tablet driver loads.
+
+When qemu mouse mode changes from relative to absolute
+we must turn off sdl relative mouse mode too.
+
+Fixes: https://bugs.launchpad.net/qemu/+bug/1703795
+Signed-off-by: Gerd Hoffmann <email address hidden>
+---
+ ui/sdl2.c | 1 +
+ 1 file changed, 1 insertion(+)
+
+diff --git a/ui/sdl2.c b/ui/sdl2.c
+index 812c315891..858e04d7c0 100644
+--- a/ui/sdl2.c
++++ b/ui/sdl2.c
+@@ -249,6 +249,7 @@ static void sdl_mouse_mode_change(Notifier *notify, void *data)
+     if (qemu_input_is_absolute()) {
+         if (!absolute_enabled) {
+             absolute_enabled = 1;
++            SDL_SetRelativeMouseMode(SDL_FALSE);
+             absolute_mouse_grab(&sdl2_console[0]);
+         }
+     } else if (absolute_enabled) {
+-- 
+2.9.3
+
+
+
+I can confirm that the patch from comment #9 appears to fix the original problem.
+
+The patch works for me too. Thanks.
+
+
+
+Fix has been included here:
+https://git.qemu.org/?p=qemu.git;a=commitdiff;h=8dfa3061ce56d871dc9df
+