Fullscreen mode

[caskolkm]

based on #235 running yosemite and amethyst 0.9.4: look at the menu bar..

Trello Card

[ianyh]

Still working for me fine. Is that Android Studio? I can test when I get a chance.

[caskolkm]

It's IntelliJ IDEA :)

[caskolkm]

look at this fullscreen mode..

[Hultner]

I get these same problems in fullscreen with all sorts of applications such as Spotify and Calendar. Also when I leave a fullscreen workspace and come back to it my window will retile so that I get a gray stripe at top where the menubar would be almost like in the first screenshot here. My solution so far have been hitting alt+t right while the the animation when going into fullscreen is still active, this usually make the application behave as expected. If there is a way for Amethyst to treat all fullscreen application as floating that might solve this problem.

[ianyh]

Calendar works for me. Haven't had a chance to test IntelliJ or Spotify.

What versions of Amethyst and OS X are you guys running?

[Hultner]

Here you can see the bug in Chrome for me.

[caskolkm]

@ianyh 0.9.4

[ianyh]

Another question: do you have displays have separate spaces turned on or off?

[Hultner]

Turned on, the bug usually appears when switching back to a space.

[caskolkm]

Turned on.

[ianyh]

I have now been able to reproduce this. Should hopefully be able to fix it soon, but I'm swamped with other stuff.

[caskolkm]

any idea when this bug will be fixed?

[ianyh]

Probably not until after the middle of this month.

[ianyh]

Spent some time looking into it. This is basically the same as #207 . The problem is that there is a race condition between changing spaces and changing the activity of windows. i.e., sometimes the space will be marked as something, but the windows in both the new space and the previous space are marked as active. Looking into ways to get around the race condition.

[ianyh]

The secondary problem is that we don't reflow windows in fullscreen application spaces, so the window gets placed incorrectly, and then has no way of getting back. I really need to figure out a way to detect when the space starts changing as opposed to when it finishes doing so.

[caskolkm]

i hope there will be a fix soon. Until now i need to quit amethyst..

[ianyh]

Having done some more digging I've actually got a fix for this in the works. I'm hoping to get a development build out this that people can test. I tried very hard to reproduce the bug with the new behavior and failed, so I'm optimistic.

[caskolkm]

let me know when devbuild is ready!

[ianyh]

@drumstick23 dev build out, let me know if it helps

[caskolkm]

Installed! In a couple of days i will report if they are still issues.

[caskolkm]

@Hultner can you installed dev build too? and check if there are still problems.

[caskolkm]

After a hour of working in 0.9.5b there are still issues, the windows size is good. The menubar is black, so the bottom of the window is not visible (falls out of the screen)

[ianyh]

@drumstick23 so the first screenshot you posted is still happening?

[Hultner]

I'm running 0.9.5b and the problems seem to still be persistent. Sorry that I didn't report it earlier, I've had a hectic couple of days.

[ianyh]

Damn. Back to the drawing board.

[caskolkm]

@ianyh not the same, in full screen mode the window is (1680 x 1050) but because it isn't placed at the top corner the bottom of the window falls of

[girvo]

I've also experienced the same bug that @drumstick23 experienced. Is there anything I can do to help debug this?

[leonardt]

I've seen this but rarely, I'll try to find a reproducible set of steps to get it to happen (@girvo that is one way to help. A list of steps that will reproduce the bug every time will aid debugging it).

However I haven't seen the ghost space for the menubar show up yet so looks good for that.

Edit: I have a feeling it may be related to full screening an app then quickly changing spaces, so perhaps it's related to the asynchronous nature of reflows?

[leonardt]

Okay, it seems I'm able to reproduce this consistently by switching between fullscreen apps using mission control (4 finger swipe up on my system).

Edit: It seems that it's not entirely caused by mission control. I think it relates to switching with mission control after you fullscreen an app that was tiled in a layout with more than one window (taking on the old shape it used to have in said layout). This makes me think that this must be related to the reflow dispatching.

[caskolkm]

@ianyh when will you fix this?

[ianyh]

@leonardt @drumstick23 I still can't reproduce anymore, so I'm not sure how to move forward. What are the specs are the machines you guys are running?

[TimSoethout]

I think you fixed the most of it, since it happens less. But sometimes it still happens on my 2014 27' imac (non 5k) i7 32gb with one extra Dell monitor with the same resolution.

[ianyh]

Is there a behavioral pattern that is triggering it? Does it happen when you first go fullscreen? Does it happen when you switch in and out of the full screen app? How do you switch in and out? Mission control? Using next/previous space navigation?

[girvo]

I think I've worked out how to trigger it, or at least what is triggering it.

So, you have a virtual desktop with no other apps on it except, say, Safari. You set Safari to fullscreen mode. This works fine, no problems! You delete or otherwise change that virtual desktop, so Safari is now "assigned" to a virtual desktop with other apps on it. Change between those virtual desktops, and viola, your full screen safari is now taking up half the screen, as if it was on the virtual desktop with the other app, even though it's full screen.

This is not the only way it's triggered, but Amethyst believing the app to be on a virtual desktop with another app despite being full screen seems to be the base symptom. Hope that helps.

[caskolkm]

good job @girvo

[ianyh]

I actually still can't reproduce on either of my machines with external displays and without.

@drumstick23 can you reproduce with @girvo's steps?

[ianyh]

Deployed another dev version with some more changes that should help prevent this from happening. Let me know if anything's changed.

[caskolkm]

@ianyh installed new dev version, i can reproduce it with @girvo 's steps.

so my story:

2 virtual desktops: 1 with chrome and iterm, 1 with hipchat put hipchat in fullscreen mode, removed the virtual desktop where hipchat was before full screen mode a little tiggering with or without mission controle, e voile, the window thinks there are 3 screens on a fullscreen window.

[caskolkm]

in my opinion what goes wrong: when a window is in fullscreen mode, there can't be placed 3 window on the screen...

[Hultner]

I'm still experiencing problems with 0.9.5b2. Some screenshots

My setup is the following 2x1920x1080 external displays and the internal 2560x1600 rMBP display running in 1680x1050 HiDPI mode. It seems to happen when I switch back to a full screen space after I've been in a "normal" tiled space. Most of my spaces run 2 tiles, sometimes 1 or 3 tiles but they're less common.

[ianyh]

Okay. There is definitely something else going on than what I thought it was. I'll have to experiment more to see if I can reproduce.

Do any of you assign apps to desktops?

[caskolkm]

what do you mean? with Amethyst?

[ianyh]

Like that.

[Hultner]

Nope, didn't know about that feature before now. Checked and all my applications appear unassigned/none.

[girvo]

No, I experience the bug without having to assign applications to specific desktops.

Sent from my iPhone

On 16 Dec 2014, at 3:40 am, Ian Ynda-Hummel notifications@github.com wrote:

Okay. There is definitely something else going on than what I thought it was. I'll have to experiment more to see if I can reproduce.

Do any of you assign apps to desktops?

— Reply to this email directly or view it on GitHub.

[caskolkm]

Lol, i also didn't know about that feature ;)

[caskolkm]

so there is a problem: i can't without work without amethyst, but i can't without fullscreen apps. When will we see some process on this bug?

[ianyh]

Unfortunately, this bug isn't linear in that spending time on it does not necessarily mean progress. I've been pretty bogged down with work and the holidays, but I'm hoping to get back to this now that things have quieted down a bit, but it's going to require a non-trivial breakthrough to figure out what's going on. I might need to prioritize #28 to get more data.

[caskolkm]

@ianyh if i can help you, let me know ( i am a developer but not related to objective C... )

[ianyh]

Okay, new theory:

Looking at the reflow calls on my machine the sequence is as follows:

• Start fullscreen
• Finish fullscreen
• Triggers space change
• Reflow
• Reflow

However, as explained in #177 , shifting GPUs can trigger reflows. If the fullscreen transition animation triggers a GPU shift you could run into the following sequence of events:

• Start fullscreen
• Reflow
• Finish fullscreen
• Trigger space change
• Reflow
• Reflow

This might explain why it happens to some people and not others, and the sequence of events would cause the observed effects. Going to see if I can figure out a way to patch things up.