Keep alive when i3bar hidden

[tobes]

By default i3bar sends a -SIGSTOP to a process sending it data. This causes
some modules interacting with i3 ipc to be suspended, which lead to locking
issues inside i3. This patch switches to using SIGUSR2 as the suspend
signal.

The async modules eg window_title_async were affected.

see #253

[tobes]

@cornerman could you just confirm that this is working for you. It should all be good but worth double checking.

[tobes]

Note: this patch just fixes the immediate issue of i3 crashing.

Further work is needed.

• We should put i3status to sleep when we are suspended.
• We should stop py3status modules too (any that need to stay alive should be using a thread).

[cornerman]

It works for me!

[cornerman]

Wait, this does not work as expected. Sorry, but I overlooked that i3status actually crashes, so we cannot just continue when we receive USR2 from i3status. There is something else going on...

[cornerman]

By default, signals are dispatched to all subprocesses and i3status exits when it receives a SIGUSR2. So, in order to avoid sending this signal to i3status, we could do this when we open the subprocess:

def preexec():
    # Ignore the SIGUSR2 signal for this subprocess
    signal.signal(signal.SIGUSR2, signal.SIG_IGN)

i3status_pipe = Popen(
                ['i3status', '-c', tmpfile.name],
                stdout=PIPE,
                stderr=PIPE,
                preexec_fn=preexec)

We could probably also translate it to SIGSTOP there.

[tobes]

Hmm this is frustrating. Yesterday I could reproduce this bug today I can't.

@cornerman is there a minimal i3status.conf that will hit this for you on master?

general {
        output_format = "i3bar"
        colors = true
        interval = 5
}


order += "static_string"
order += "load"
order += "window_title_async"


static_string {
    format = "hello world"
}

window_title_async {
    always_show = True
}

feels like it should be enough but I can't trigger stuff

[cornerman]

is there a minimal i3status.conf that will hit this for you on master?

Seems that it happens with any config for me. What happens if you do:

killall -s USR2 i3status

For me, all i3status processes exit and py3status busy waits as we now ignore that i3status exited.

[cornerman]

This will also fix #179

[tobes]

OK, I can trigger it again via i3-msg bar mode hide and i3-msg bar mode dock

Should have an updated patch soon. Thanks

[tobes]

@cornerman could you test the latest version which includes your fix and also should stop i3status (which should wake up too)

[cornerman]

Nice, it works!

[tobes]

Nice bit of teamwork there 👍

@ultrabug this patch should be ready for consideration now

[ultrabug]

sorry if it should be obvious, but why negate the SIGUSR2 value here ?

[tobes]

That was an odd thing but is no longer needed as we do not send the SIGUSR2 to i3status now. Maybe an i3status bug?

[ultrabug]

You guys baffle me 👀 quick question and I'll merge. Thanks

[ultrabug]

why not self.i3status_pipe here directly ?

[tobes]

I've moved this into the Popen def is that what you meant?

[tobes]

I've updated this should be better now

[ultrabug]

Thank you, yes.