Page 27 of 71

Re: POTZBLITZ V5.2 (beta), OCT 5th

Posted: Mon Oct 09, 2017 7:36 pm
by KOB001
The rail_travel.lua is used for railroad-pathfinding. So, what were the changes you made, forbid rail movement for the AI entirely? That of course would severely weaken the AI.

Are you experienced with lua? Any help would be most useful as four eyes see more than two, especially when it comes down to fix the replay crash bug, which should have top priority
.
Only a little but the more eyes the better.

Hey if I could forbid rail movement it would at least show the file did something!
The hosting player is meant to start the Potzblitz t 1914-1918 scenario.
Any other scenario is guaranteed to crash or behave weirdly (even more than Potzblitz) due to the many changes made to the scripts as detailed in the opening post.
Hmmm Zombo hosted a 1914 game.

Could the source of the replay bug be something as simple as that?

I agree the MP bug should be top priority.

When exactly did the replay bug start showing itself (which versions etc)? Has anyone managed to complete a game without the bug showing up? If we had a list of all reported MP crashes we might have better luck identifying a source?

Re: POTZBLITZ V5.2 (beta), OCT 5th

Posted: Mon Oct 09, 2017 7:48 pm
by Robotron
KOB001 wrote:When exactly did the replay bug start showing itself (which versions etc)? Has anyone managed to complete a game without the bug showing up? If we had a list of all reported MP crashes we might have better luck identifying a source?
Zombo should know best as of my understanding he managed to complete several Potzblitz MP matches of earlier versions.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Mon Oct 09, 2017 8:07 pm
by Zombo
Could the source of the replay bug be something as simple as that?
I'm afraid not, the games I played vs Robotron were hosted by him and terminated by the infamous replay crash
Zombo should know best as of my understanding he managed to complete several Potzblitz MP matches of earlier versions.
the last MP match I completed was using 3.4x, if I'm not mistaken, and it never happened then. Double retreats never happened then either, don't know if it's related in any way.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Mon Oct 09, 2017 9:34 pm
by Robotron
The source of the bug must still be lingering within the combat script then.

Thing is: the original combat script contained NO random-number-functions and it seems for a very good reason: namely to assure that the exact same values are guaranteed to get passed over from the current player's combat phase to the other player's replay phase. For the same reason it is virtually impossible to get different combat results from reloading for any given battle in the vanilla game. There's no fluctuation because of the absence of randomness.

Being a fan of rolling LOTS of dice I did of course include a plethora of additional random modifiers to fire during the combat phase hoping the results of the die rolls would be passed over in exact the same sequence.
This is undoubtedly NOT the case as double retreats, wrong hitpoints numbers etc. have been observed, leading to a botched up unit array.

So my next/final attempt will be to bite the bullet and exchange all the random stuff I added to the combat script for fixed values.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Mon Oct 09, 2017 10:13 pm
by KOB001
Robotron wrote:o my next/final attempt will be to bite the bullet and exchange all the random stuff I added to the combat script for fixed values.
Do you have the old values from 3.4 still?

Presumably they would give you a base to work with (changes could be highlighted easily with the 'compare' plugin on Notepad++)?

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Tue Oct 10, 2017 2:17 am
by KOB001
Found that Airship crash I mentioned a while back, it's attached.

However, so far, it's been the only crash of my latest "Warplan East" attempt (this time on normal difficulty).

I also appear to have successfully disabled AI rail moves... interestingly, although I've noticed the lack of a rail move hurt the AI on one occasion so far, the benefits to the AI of this disabling FAR outweigh that one incident.

For instance the AI are forming much more stable defence lines.

Also most of their units are in the white, most of the time... in fact, other than due to my artillery fire, I've not seen a singe AI unit in the red so far!

They also don't seem to actually move artillery or aircraft onto the front either... instead that AI mishap appears to be an artifact of their rail behavior.

The AI generally seem so much tougher (I got NOWHERE near Riga this time) that I've just basically resorted to ridiculously massive artillery barrages on my Eastern front.

So, if this keeps up (I'm approaching 1916), perhaps one of the best ways of improving the AI (over the course of the war) would be to disable its rail moves & instead provide it a few protective spawning events for unexpected amphibious invasions/opening of new fronts?

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Tue Oct 10, 2017 11:29 am
by Robotron
KOB001 wrote:I also appear to have successfully disabled AI rail moves...
Upload the script please so I may verify.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Tue Oct 10, 2017 11:54 pm
by KOB001
Robotron wrote:Upload the script please so I may verify.
I was verifying for myself. I think I saw one rail move, so rail moves might not be fully disabled. Either way, the AI seemed far tougher than I'm used to and I only saw the AI move a support unit onto the front line once (and even then it withdrew it the next turn).

The test game ended with repeated crashes during the 1916/17 winter (I think due to "badharvest" errors).

All crashes attached (only crash 3 repeated... others continued to work after reloading).

As for how the game was going Russia, despite my deluge of artillery, still hadn't surrendered but it was starting to crack. The Salonika Front looked good for me though & Romania had joined the CP. I didn't notice Cardona on the Italian front & I was surprised by French units being there (were they shipping them around the impassable hexes?).

Turkey hadn't joined the war yet (in 1917!) & despite me not invading any Neutrals or attacking any merchants the US were still about to join the war! With AH looking very shaky & the blockade causing pain I'm not so certain I could've won.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Tue Oct 10, 2017 11:56 pm
by KOB001
Robotron wrote:Upload the script please so I may verify.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Wed Oct 11, 2017 12:10 am
by KOB001
As for the MP game with Zombo (using the 1914 scenario admittedly) that ended in a replay crash.

Issue does appear to be combat related.

I saw his Naval attacks, then his attacks on the Adriatic coast, then he swapped some Serb units & then it crashed.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Wed Oct 11, 2017 3:51 pm
by KOB001
Robotron wrote:Upload the script please so I may verify.
I just ran through a quick Entente game... I'm more than a little confused if that railmove script is doing anything at all.

For 1914 the AI CP were zipping units around & moving artillery onto the front line as normal.

Then in 1915 & 16 the AI stopped zipping around, stabilized its front lines, looked after the efficiency of its units & avoiding placing support units on the front lines.

In late 1916 the AH surrendered due to hunger & Germany started zipping units around again before a repeating crash ended the game.

Either way I've attached the bug.

The only other thing I noticed was that there's an event called "warpropaganda2" with no text description.

The game itself was a bizarre one, Italy joined the CP despite them not doing particularly well. The US joined early (again) & were poised to launch numerous Mediterranean attacks on Italian islands & colonies. Belgium surrendered in 1915, threw out all belligerents & trapped a British artillery transport (I'm assuming that's not meant to happen!). Bulgaria & Romania were both staying neutral & Greece also seemed to refuse to get involved & gift me Salonika.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Thu Oct 12, 2017 1:03 pm
by Robotron
Koob, thanks again SO much for your dedication to flush out bugs but alas, this time there were several crashes without error message in the logfiles which, as you will already know by now, are almost impossible to fix. I rest my hope upon that those are caused by the defective combat script as I detailed earlier which I'm still rewriting.

As for the railroad-madness I see my notion confirmed by your efforts that the corresponding scripts while getting checked are overruled by the core game.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Thu Oct 12, 2017 3:56 pm
by KOB001
Robotron wrote:As for the railroad-madness I see my notion confirmed by your efforts that the corresponding scripts while getting checked are overruled by the core game.
Seems it.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Fri Oct 13, 2017 8:35 am
by Zombo
KOOB, thanks too for your contribution

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Sat Oct 14, 2017 8:40 pm
by Robotron
Testing V5.3 now, all seems stable.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Sun Oct 15, 2017 4:11 am
by KOB001
Robotron wrote:Testing V5.3 now, all seems stable.
Is 5.3 ready enough to test the MP?

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Sun Oct 15, 2017 9:38 am
by Robotron
@KOB: Not yet. It occurred to me that the problem of different results from random die rolls happening in the first player's COMBAT phase and the second player's REPLAY "phase" might also cause trouble when certain damage-causing events randomly trigger while MOVING units.
This would mean I'd have to rewrite some events too if my assumption was correct.

It would be very helpful to me if Zombo could remember whether he was able to finish a game where the "Sea Mines" event was already included. Sea Mines were included in V3.0.
If the replay crash bug did not yet happen there it would save me an awful lot of work and I could upload the new version right away.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Sun Oct 15, 2017 2:53 pm
by Robotron
A small success...sort of...to at least help the dilemma of having too many unfinished and therefor "lost" games clogging up available game slots because of the replay crash bug.

For V.53 I added a "surrender" button that allows the player who is suffering from the replay crash bug - and therefore is unable to continue the game - to surrender the game from the MP panel.

Although the game then crashes, upon reloading said game is flagged as "over" and thus can be removed without the need to contact Slitherene staff and wait for them to remove the offending game.
klappts.jpg
klappts.jpg (9.15 KiB) Viewed 1292 times

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Mon Oct 16, 2017 10:00 am
by Zombo
It would be very helpful to me if Zombo could remember whether he was able to finish a game where the "Sea Mines" event was already included. Sea Mines were included in V3.0.
I would say no, not finished. Gone as far as 1916.

Re: POTZBLITZ V5.21 (beta), OCT 8th

Posted: Wed Oct 18, 2017 3:56 am
by operating
Robotron wrote:A small success...sort of...to at least help the dilemma of having too many unfinished and therefor "lost" games clogging up available game slots because of the replay crash bug.

For V.53 I added a "surrender" button that allows the player who is suffering from the replay crash bug - and therefore is unable to continue the game - to surrender the game from the MP panel.

Although the game then crashes, upon reloading said game is flagged as "over" and thus can be removed without the need to contact Slitherene staff and wait for them to remove the offending game.
klappts.jpg
Nice move with the new button...!!!! :wink: