On 2022/06/09 21:02, Gerd Hoffmann wrote:
On Thu, Jun 09, 2022 at 08:45:41PM +0900, Akihiko Odaki wrote:
On 2022/06/09 19:28, Gerd Hoffmann wrote:
--- a/include/ui/console.h
+++ b/include/ui/console.h
@@ -139,6 +139,7 @@ typedef struct QemuUIInfo {
int yoff;
uint32_t width;
uint32_t height;
+ uint32_t refresh_rate;
} QemuUIInfo;
/* cursor data format is 32bit RGBA */
@@ -426,7 +427,6 @@ typedef struct GraphicHwOps {
void (*gfx_update)(void *opaque);
bool gfx_update_async; /* if true, calls graphic_hw_update_done() */
void (*text_update)(void *opaque, console_ch_t *text);
- void (*update_interval)(void *opaque, uint64_t interval);
void (*ui_info)(void *opaque, uint32_t head, QemuUIInfo *info);
void (*gl_block)(void *opaque, bool block);
} GraphicHwOps;
So you are dropping update_interval, which isn't mentioned in the commit
message at all. Also this patch is rather big. I'd suggest:
(1) add refresh_rate
(2) update users one by one
(3) finally drop update_interval when no user is left.
thanks,
Gerd
I think 1 and 3 should have to be done once since refresh_rate and
update_interval would interfere with each other otherwise.
Well, between 1 and 3 both old and new API are active. Shouldn't be
much of a problem because the GraphicHwOps implementations are using
only the one or the other.
take care,
Gerd
The only GraphicHwOps implementation updated with this change is xenfb.
xenfb can be switched to use refresh_rate in step 1 or 3.
Switching to use refresh_rate in step 1 would break the refresh rate
propagation until all host displays are updated to set refresh_rate
instead of calling update_interval.
Switching to use refresh_rate in step 3 would break the refresh rate
propagation when a host display is updated to set refresh_rate instead
of calling update_interval but xenfb does not use refresh_rate.
Regards,
Akihiko Odaki
