wlserver: Re-hook pausing on session pause

[samueldr]

During the refactor in 88eb1b4, the handle_session_active function lost the ultimate role of causing a pause when the session became inactive.

The duty of pausing the session was given the DirtyState function on the backend, which now uses the same "moral" condition to set the paused state (g_DRM.paused = !wlsession_active();)...

... except that now DirtyState state is only called when the session is resumed. In turn, this means that on session suspend, nothing ends-up pausing the DRM backend anymore!

This change unconditionally calls DirtyState, which in turn does the accounting for pausing the backend. Actually, it conditionally passes false to the argument to force nothing.

This fixes what ends-up causing drmModeAtomicCommit: Permission denied when moving to another VT from gamescope's.

---

Hi again! 👋

Hot off the heels of #1584 I was writing-up a bug report to solve the cause of the symptom I was observing.

As it turns out, while showing where my knowledge about the situation ended, I realized the solution was within the scope of my knowledge.

I'm not 100% sure of the actual current implementation (passing the active flag to DirtyState), there may be something more apt to do nowadays to actually pause the renderer, but looking at the change that AFAIUI introduced this regression, this is recovering the initial intent.

Anyways, the fact all the hard work was already done made it trivial to add the feature to “Handle seat disable/resume for VT switching”.

Don't hesitate to discuss changes, or even directly send them here if anything trivial.

How this was tested

Manually, using a Jovian NixOS system (SteamOS UI and system semantics) on my Steam Deck. It was on top of gamescope 3.15.11, but it shouldn't really matter compared to the current revision on master.

I'm using the symptom (the libliftoff message) as a success metric.

Going from:

[gamescope] [Info]  wlserver: [libseat] [libseat/backend/logind.c:382] Disabling seat                                                                       
[gamescope] [Info]  wlserver: Session paused                                                                                                                
drmModeAtomicCommit: Permission denied
drmModeAtomicCommit: Permission denied
< snipped the many repeats >
drmModeAtomicCommit: Permission denied
drmModeAtomicCommit: Permission denied
[gamescope] [Info]  wlserver: [libseat] [libseat/backend/logind.c:379] Enabling seat                                                                        
[gamescope] [Info]  wlserver: Session resumed

To:

[gamescope] [Info]  wlserver: [libseat] [libseat/backend/logind.c:382] Disabling seat                                                                       
[gamescope] [Info]  wlserver: Session paused                                                                                                                
[gamescope] [Info]  drm: Connector eDP-1 -> VLV - ANX7530 U            
[gamescope] [Info]  drm: [colorimetry]: EDID with colorimetry detected. Using it                                                                            
[gamescope] [Info]  drm: [colorimetry]: r 0.684570 0.314453                                               
[gamescope] [Info]  drm: [colorimetry]: g 0.245117 0.714844                                                    
[gamescope] [Info]  drm: [colorimetry]: b 0.137695 0.049805                                          
[gamescope] [Info]  drm: [colorimetry]: w 0.312500 0.329102                                               
[gamescope] [Info]  wlserver: [libseat] [libseat/backend/logind.c:379] Enabling seat
[gamescope] [Info]  wlserver: Session resumed                                                                                                               

Note that there are messages that are being shown on pause that wouldn't beforehand. I believe it may come from my "misuse" of simply calling DirtyState. That might be inconsequential in the end.

[samueldr]

(Ooops, looks like I'd referred to the wrong commit in the message, last force push is only fixing the commit message.)