Hotfix build 79600 is live!

Ok, so they may still implement it yet, that's good to hear. Though it would be nice if they at least split it up, just for modders

Vanilla comfabs in 79317 can't assist bot factories, so perhaps it was fixed.

This build has added UI bugs on my Linux system, and there have been server issues in every game I've played.

The UI will sometimes freeze. When it does, I can't click anything and the keyboard is disabled, but I can still see units moving around and icons being drawn correctly. Every time it's happened, it's been while I was navigating a fabber build menu with the mouse. I might also have been holding shift and panning with wasd at the same time. In the frozen state, I could see multiple build icons lit up as my mouse had been moving between them. It might be related to the new wall icon, as I think I had just moused over the wall each time.

I've also noticed that post-game chronocam is slightly broken. Eco figures freeze up (though UI isn't frozen) and selecting a player doesn't limit me to that player's view. But the replay itself is still accurate.

I did have some mods installed before I deleted my old PA and pamm folder. I'll make sure they're as uninstalled as pamm claims, and then I'll report the bug on the tracker.

Also, every game has involved either short (1-2) sec or permanent freezes of movement, typical of connection problems. Everyone in the server experienced it at the same time.

I run Arch on a Thinkpad T420 with an Intel HD3000.

Hi everyone, I'm new here but couldn't help but share this.

I found this article a while ago on hacker news and it could be applicable to the pathfinding algorithm with regards to catching up to a moving target and might be worth looking into.
(I'd link to the article but can't as I'm new)
It discusses missile guidance systems (which face essentially the same issue as trying to catch up with a group) and finishes with a discussion of the "optimal" way for something to catch something else, which is proportional navigation (insert link to wikipedia here)

Hm. That's a very good point (I have no experience writing pathfinding or AI algorithms)

Wow, now that I think about it some more, that's a very unpleasant problem when you take real issues into account.

Maybe you could cast rays from the unit trying to join the group to the edges/corners of the different obstacles surrounding the unit in 360 degrees (since you'll always want to move in a straight line) and from each of those points, consider the problem again after moving the units that you are trying to join along their path by the amount of time it takes you to move to the new position. Then when you've reached a point from which PN can meet up with them do that from there on.

Something like:

Code:

let join units be a method on a unit object, self, with some variables defined as convenient,
and let unit_path by the path of the units that self is trying to join.
join_units(unit_path):
    possible_paths = PriorityQueue()
    # get a circle of obstacles from where we currently are to search for path starting points
    for obstacle in self.get_obstacles_in_circle():
            for corner in obstacle.corners_from(self.position):
                possible_paths.append(Path(self.position, corner))
    #take the path that has the least euclidean distance to unit_path ala A*
    while(True):
        with possible_path as possible_paths.pop():
            if possible_path.can_reach_in_time_at_speed(unit_path, self.speed):
                return possible_path
            # do the same thing where we append corners from the current path's view of obstacles around it as paths to the possible_paths queue

I wanna see what happens when he sits down on that.
*BZZZWWM* *slices fall through*

Is this with the --software-ui switch enabled? If not try launching pa via console and add that tag in. It disables hardware acceleration for the ui and usually fixes freeze problems.

Note there is a bug with software ui mode- you'll get stuck at a black screen post game instead of going back to the Menu. Switching between windowed and full screen fixes it apparently, though the keyboard combo for that (alt Enter) doesn't work in ubuntu so haven't been able to try it myself.

Nice to finally see a new stable build! But could you explain how the different types of orbital radars now work?

They haven't changed in behaviour, only ranges.

Sorry, I meant to get back on you on that. That keyboard setting is in settings -> keyboard -> general -> toggle full screen. Mine is set by default to <Alt>+<Enter> (and works fine under Kubuntu). Maybe it works better for you if you change the shortcut. (If playing in Windowed mode, resizing the window also does the trick.)

Deepspace radar now shows strategic blips instead of full orbital vision.

Orbital units have increased orbital vision, so orbital scouting works a little more like regular surface units.

auto attack is very buggy.

Lots of units not attacking other units which are in range.

Also noticed that you can still slow down bombers drastically as they attack. It's very easy to do and gives an enormous advantage to snipes.

That algorithm isn't applicable - since you don't have an unit_path in PA. For performance reasons, PA doesn't run A* for each individual units, but instead integrates terrain cost into a generic potential field for each current goal, which is then used by multiple units to navigate into the direction of lowest potential until all units have reached proximity of the goal. Roughly the same computational cost as running A* once, at the cost of being unable to release the memory allocated by the A* run until all units have discarded the goal.

It would be possible to adapt that algorithm to potential fields, but it would be quite costly and couldn't account for slow downs of the target group due to flow capacity limits around obstacles, so it might even overshoot unless you recompute it frequently.

@jables

On OS X the game is often unplayable for me now during game start with the following:

• slow start with lots of Requesting resource read for [filename] with internal id [id] reported FAIL slowly scrolling up the console (may not be cause but it's all I see in the console logs)
• when it does start panels often fail to load requiring another refresh or restart and more delay

This is what I was seeing in PTE. When it happens it's really bad.

By the time I actually start, refresh and/or restart PA the game has started, my com has auto landed somewhere and everyone else is well underway.

Reproduced with --nomods.

These also appear in logs:

• Argument conversion failed: Wrong type - expected Integer, got Null while converting argument 0 for handler panel.messageById

Anyone else seeing this on OS X?

Hmm, I just had the losing-the-ability-to-select-and-click-on-stuff bug under Linux with this build. I can't remember the last time that happened to me. Many months ago anyway.

I have dozens of these things in the log by the way:

Code:

[09:26:12.643] INFO [COUI] PID: 9413 | 9413 13:26:12.596639 [9413:9413:ERROR:render_widget_host_view_gtk_coherent.cc(289)] Unable to get backing store!

But those appear in the other logs as well, so probably not related.

I use the --software-ui flag in all of my games, in case that's relevant.

Another thing is you can assist factories which are turned off.

Say i have a bunch of stuff assisting a t2 factory but i want to power it off, it'll keep building and using up eco.

After matches Chronocam is buggy, it freezes at the final eco rather than showing the current eco of the bit you are up to. not sure if this applies during the game.

This is the same as it was in SupCom, and it's annoying as hell.

Can confirm. This happened to me yesterday while I was reviewing a match for my build order.
It also happens during replays from what I've heard.