Page 1 of 1

My testing for 0.4.6 .

Posted: Sun Apr 30, 2017 9:06 am
by danieledel
I was testing the new Freeorion 0.4.6 on Linux Lubuntu, and it works fine. Very useful for those NOT familiar with English can copy and paste in Google Translator the lyrics that are in Pedia. Playful, I'm better off with the technology tree.

A note to laugh: how ever one of the constructible structures on the planets is a pyramid with inside an eye? Are you funded by the Illuminati? Ahahahahahahaha!
Good day.

Re: My testing for 0.4.6 .

Posted: Sun Apr 30, 2017 3:27 pm
by defaultuser
Doesn't seem too odd to us 'Muricans:

https://en.wikipedia.org/wiki/Eye_of_Providence

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 9:16 am
by danieledel
In Europe is a symbol associated with the population of Freemasonry and the secret groups, is rarely considered the symbol of God.

I'm playing with Freeorion version 0.4.6. If a new 19 star galaxy is active i always see the same map, as if randomness was blocked. I have also appeared unnoticed monsters, but nothing bad about it. In order to have a different map i have to change the number of stars, is it also for you? Do not change the shape of the galaxy, maybe should i try this option? In version 0.4.5iI changed the map when i started new games, even though the number of stars and the shape of the galaxy were the same. Good day.

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 9:50 am
by Geoff the Medio
danieledel wrote:If a new 19 star galaxy is active i always see the same map, as if randomness was blocked.
Are you using different random seeds?

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 9:52 am
by MatGB
You should always get the same map if you use the same seed, that's deliberate, if you hit the dice symbol next to the seed you'll get a different one.

(and it's my understanding the Eye became a symbol of Freemasons after the USians adopted it for their currency but given the secretive nature of some of the organisations it may've been around before)

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 10:06 am
by LGM-Doyle
In the single player galaxy setup the first box is "Seed" and beside it is a dice icon.
The random seed is used to randomize the galaxy generation, without changing the other settings.

To get a new random galaxy either:
  • Click on the dice icon.
  • Type in any random sequence of text

The seed feature is useful for testing because typing in the same seed creates the same galaxy and allows a dev to reliably look at the same bug.

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 5:40 pm
by defaultuser
LGM-Doyle wrote:The seed feature is useful for testing because typing in the same seed creates the same galaxy and allows a dev to reliably look at the same bug.
Unless things changed for 0.4.7, it's only the star systems (excepting the homeworld) that are the same. Monster and native species placement and such will still vary.

Re: My testing for 0.4.6 .

Posted: Tue May 09, 2017 6:26 pm
by MatGB
defaultuser wrote:
LGM-Doyle wrote:The seed feature is useful for testing because typing in the same seed creates the same galaxy and allows a dev to reliably look at the same bug.
Unless things changed for 0.4.7, it's only the star systems (excepting the homeworld) that are the same. Monster and native species placement and such will still vary.
Monsters vary, including guard monsters, but everything else should be the same. It's not sometimes and there have been bugs.

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 12:40 am
by defaultuser
MatGB wrote:Monsters vary, including guard monsters, but everything else should be the same. It's not sometimes and there have been bugs.
Does that include natives? I'm pretty sure I've seen them vary in the past. I don't think I've specials vary, other than their guards. I don't do a lot of restart from seed though.

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 3:13 pm
by MatGB
It should include natives, there have been glitches, if you see it again (if/when we ever get 0.4.7 running for you) then please try to report it as best as possible, not many of us do replicate deliberatel that often, but when we need to we really need to and native placement is something I'm sometimes test (specifically if trying to test a new native species, the Lembala'Lam were a pain to test due to their quirks).

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 4:22 pm
by Oberlus
it's only the star systems (excepting the homeworld) that are the same
Shouldn't RNG's seed determine all the random effects of the game?
I mean, is there any reason to use any other source of randomness when testing games with a given seed?
I understand that in real games (specially multiplayer ones) random effects shouldn't be predictable to avoid potential cheating, but once we are passing a given and known seed to the game, why not use that one and nothing else?

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 5:59 pm
by Dilvish
Oberlus wrote:
it's only the star systems (excepting the homeworld) that are the same
Shouldn't RNG's seed determine all the random effects of the game?
I mean, is there any reason to use any other source of randomness when testing games with a given seed?
Under typical settings there are a couple additional ways that some extra variation can creep in slowly, stemming from having multiple threads and processes, and variability in the relative timing of server requests made by those asynchornous event sources. To avoid that you would need to play with just a single AI, and also go into options/misc and change the number of effects processing threads to 1, and then I believe everything should remain fully repeatable.


**edit-- it just sunk in that the quote was about the galaxy creation stage-- for that it is probably enough that there is just one effects processing thread, regardless of how many AIs. Also we have had a hard time eliminating a small bit of variance between the results on different OS's, although it's possible that we recently tracked down the last source of that, from the use of unordered sets, although I am not 100% recalling if that got fully resolved.

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 6:58 pm
by Oberlus
Ah, I understand then.
Dilvish wrote:Under typical settings there are a couple additional ways that some extra variation can creep in slowly, stemming from having multiple threads and processes, and variability in the relative timing of server requests made by those asynchornous event sources. To avoid that you would need to play with just a single AI, and also go into options/misc and change the number of effects processing threads to 1, and then I believe everything should remain fully repeatable.
Regarding gameplay of AIs during the whole game, I guess there are ways to use a random number sequence that is independent of concurrent processing, like using a different seed (that depends on the initial one) for each AI thread and passing the last generated number to the following call of the generator. Not sure if this is easily available on C++11, but it is possible in plain C.

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 8:06 pm
by defaultuser
MatGB wrote:It should include natives, there have been glitches, if you see it again (if/when we ever get 0.4.7 running for you) then please try to report it as best as possible
It's quite possible I'm misremembering.

Re: My testing for 0.4.6 .

Posted: Thu May 11, 2017 8:30 pm
by Dilvish
Oberlus wrote:Regarding gameplay of AIs during the whole game, I guess there are ways to use a random number sequence that is independent of concurrent processing, like using a different seed (that depends on the initial one) for each AI thread and passing the last generated number to the following call of the generator.
I don't think that's the issue-- I am pretty sure that the AI's already do not share the the same RNG (and side note, the state in the RNGs we use is far more complex than simply what was the last number generated). The issue is that the AI's, running in independent threads, submit various requests to the server in an order which depends on how the CPU core threading is being handled, which is opaque to and uncontrollable by us. In particular, I think it is the requests which create/destroy objects (splitting or destroying fleets, or destroying ships or buildings, I suppose), whose varying order can wind up making a difference in later results from the server. I don't recall the exact details though. For some major events like combat we reseed the RNG right before the combat resolution to make sure that the result does not depend upon such vagaries, but it's my recollection we currently don't reseed before every use of the server RNG and anyways I think that the varying order of object IDs can wind up changing the significance of the RNG results even if the RNG number stream remained the same. It's been a while since I've really looked closely at that, just take this as some tips if it was something you really wanted to dig into the code.