Using macos client, xpra v3.0.2-r24371 python2.7 64bit on macos 10.13.6 (High Sierra) and server version on centos-6, xpra v1.0.14-r24088 (xpra.x86_64 1.0.14-1.r24088.el6_6).
If connected client had been iconified, and I wake up machine from sleep (machine is maintaining network connections during sleep, it's a desktop not a laptop), the client immediately de-iconifies (1).
I then iconify it again, to start working in unrelated windows. But typically the client again de-iconifies about 20 seconds later (2).
My expectation is like other well-behaved applications: remain iconified, not open triggered by wake from sleep; and once reiconified, don't auto reopen short time later.
I don't recall this behavior in past releases, but I may have just been much better about always exiting the client before sleep.
This is a minor annoyance only.
sto6: can you please run the client from the command line and add -d metadata,state,osx
and post the log output of when you resume?
I only have virtual machines of macos at the moment, and those don't sleep.
debug log of sleep/resume
Same client/server as above. Attached resume20191122a.log
Machine slept for about 1 minute. Log shows two lines: "system resumed, was suspended for". I control-C-ed the 'xpra attach' shortly after 2nd resume.
When remote desktop reopened 1st time, it was showing spinners (and wrote "drawing spinners" line); I don't recall that behavior (spinners on first auto-open) but maybe I wasn't paying attention (just re-iconifying it), anyway I suspect it's not material to this ticket.
The first de-iconification is because we are forced to re-initialize the windows after we resume from sleep to clear the window pixel buffer, this was added back in r10573 (see #492). I'm not sure if we can do this re-initialization and preserve the minimized state of the window. Will check.
The second one is very likely to already be fixed in r24466. (#2482) You can find beta builds with this fix already applied.
Don't you get disconnected from the server when you suspend and resume? (connections should be dropped if there is no response for 60 seconds)
The suspend + resume deiconification can be tested more easily using r25167.
Only macos has this bug. The GTK docs state: Gtk.Window.iconify: It’s permitted to call this function before showing a window, in which case the window will be iconified before it ever appears onscreen.
But this is not honoured on macos, where we get a map event.
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/2484