bridgewater-brunel.me.uk - Simutrans-Experimental - Pak128.Britain-Ex 0.9.1

[Sarlock]

Of course this begs the question as to why the game didn't crash in the first four season changes and only now in 1751.

[jamespetts]

Thank you all for your feedback on this. The threaded season code has not made it into Experimental, and in any event the server has only a single core, so is compiled as a single threaded application, so it cannot be this. I will try what Sarlock recommends to attempt to narrow the trouble.

[ӔO]

So I ran the game locally until 1805, and even in this 1/4 minimap image, you can sorta see how the cities really grow without any player help.


And by 1805 there were 1159 industries with a global population of 2,829,234

[Sarlock]

Wow... look at that megalopolis on the long inlet on the north end of the eastern island.  It practically covers the entire northern coast.  I'm serving about half of the island with a break-even passenger/mail network with the intention of big profits in 25-50 years due to population growth.  Looks like it'll pay off.

[AP]

I do find myself wondering if the map is a bit too crowded. Because of all the sea, and the settlements-like-water setting , there are a lot of settlements very close together. I could see it becoming a very urban map later on.

[ӔO]

There is plenty of shallow water for land reclamation in certain areas, which is probably cheaper than demolishing through built up cities.

[Sarlock]

The edge islands themselves are massive.  Once land transport becomes feasible there is a lot of potential for vast rail networks to service the western and eastern islands.  Both islands have a lot of cities locked up in their interiors.

All of the eastern islands are connected by shallow water, as are all of the western ones.

Given that land transport isn't really available in any usable form for another 50-100 years, we're stuck with sea transport either way.

[AP]

What is it that constrains passenger journeys now, is it journey time? Would there be any passenger demand for the months-long voyage between islands?

[ӔO]

Hard to say, but I don't think there is going to be much demand for those voyages at this point in time.

In another test game, pax hardly wanted to use the 27km/h train. With 80km/h max trains, I finally saw 12 carriage trains that were full. Both were not profitable, btw. Probably needed a larger network for more frequency and more demand.

[Sarlock]

I've set up a network of about 40 cities connected by ships and it's barely covering the operating costs (it may even be losing money when maintenance costs are factored in - my freight operations are likely subsidizing my passenger network).  All of the feeder lines are money losing and it's just the main trunk lines that produce moderate profits.  The ships between the hubs are running with 50-100 passengers per trip on average - enough for a small profit but nothing substantial.  The vast majority of prospective passengers decide that the journey time is too slow to want to bother.

In past games, as soon as steam ships appear passenger counts rise significantly.  The additional speed brings a lot of new customers online.  In other past games I've also been successful with inter-island transport - but you need to connect enough cities on either end to be able to produce enough passengers for the journey.  I don't have any long distance inter-island ship lines yet so I can't say if this will work in this game.  My longest passenger journey is about 6 game hours (~800 tiles=100km) distant and passengers are happy to take this (with no refunds).

It's interesting to note that while the passenger depots of the era are only 3 tiles in radius, passengers are willing to walk from much further out to reach the station, so effective coverage of the dockyards is sufficient to serve the bulk of the citizens in each city.

Another thing I'd like to share: You have a cargo ship that arrives at a dock full of cargo and loads it up, 225t and it's full.  But now it waits for 2 hours for the transfer time.  If you jump to the next location in the schedule, the ship will happily disembark and move on to its next destination with the holds nicely full of cargo without having to wait for the 2 hours to load.  It seems the cargo loading/unloading process is cosmetic only and the actual transfer is done immediately upon arrival.  Good way to save 2 hours, especially in the vital first few months when you want your first profits to roll in.

[ӔO]

btw, it is historically realistic that freight is the profitable side of business, as railway lines mainly sprouted up to carry coal from mines to factories and cities as their main source of power.

[AP]

The problem we have with sea travel and historical realism though is the inability of ships to carry different cargo.

Historically there were "triangles of trade" where a ship would carry 3 very different cargoes in a triangular route, or even just a different cargo 'out' than 'back'. I.e. fully loaded 100% of the time .In Simutrans there are very few opportunities for this*, so when ships and vehicles are created and priced based on realistic data, they will inevitably make less money as a consequence. Because we have to "pick a freight type".


*Coal/Stone/Clay works (all being bulk good). Much later, logs/steel works (both long goods). Maybe fish/dairy (cooled goods). Piece goods may work but the quantities are small.

In theory you could do coal in both directions on a route (ColliaryA to MerchantB and Colliary B to MerchantA) but normally the suppliers are linked to the local consumers, so that ends up not working profitably.

[Sarlock]

A great point, AP.  I've had the same thought.  pak.Britain isn't too bad as there isn't that many freight types, but pak128 is particularly bad in that there are too few opportunities for freight backhauls due to the restriction to certain goods types.  Keeping the number of freight types low is important in this respect.

I haven't given it much thought yet, but with such a large map there must be opportunity for significant backhauls.  That's my next area of focus.

[asaphxiix]

I'm making good money now with land transport on this map using stage coaches with a stop in every town. It really is possible.

[AP]

I haven't given it much thought yet, but with such a large map there must be opportunity for significant backhauls.  That's my next area of focus.

Fewer than you'd think between the two main archipelagos:

Bulk goods: I could only find two bulk goods wanting to go west to east (stone) which I am using to balance the coal going east to west. No potteries on the eastern island, so clay isn't required there.[Edit: and no demand for woodchips due to brewery issue]

Chilled goods: fish/meat/milk?  In 1750 it would seem implausible if we can start shipping that stuff too far... assume the speedbonus would make it uneconomic??

Piece goods: Beer / vegetables / furniture / clothes (can't find a textile mill though) . Only in small quantities.

Long goods: only timber in play at the moment.

As I wrote before, there may be a lot of industries, but given the size of the map, there actually aren't that many.

I'm making good money now with land transport on this map using stage coaches with a stop in every town. It really is possible.

How - the server game isn't fixed yet??

[ӔO]

Wrought iron is long goods. It's not particularly profitable, but it is not the worst either.

Milk and Planks are the most profitable.

In fact, I am half tempted to rename my company "Sea Cow" and pick up more milk chains.



oh, and if anyone is wondering, I won't be doing any pax until rail comes about.

[asaphxiix]

locally...

[jamespetts]

Returning for a moment to the bug, deleting Norley Wood does not help: the error merely manifests in a different building. Investigating further...

Edit: Reverting Neroden's changes to dingliste.cc have no effect: exactly the same error occurs.

Edit 2: Using Pak128.Britain-Ex 0.9.0 instead of 0.9.1 makes no difference.

[AP]

Milk and Planks are the most profitable. In fact, I am half tempted to rename my company "Sea Cow" and pick up more milk chains

Scanning the islands, there are numerous rather absurd 'opportunities' for long distance milk shipping routes. Indeed the industries are sufficiently-non-interlinked that I could run milk ships full both ways selling milk from east island to the dairies on the west island and vice versa. Presumably the consumers are very discerning and the milk is UHT...

Being 1750, dairies are one of the few industries we have a lot of! (and they keep spawning...)

[jamespetts]

After several hours of testing, I can confirm that the crashing on the May/June month boundary does not occur with the previous saved game in 2029. This is, I have to say, very mysterious indeed, and I am struggling to understand what the current saved game has that the previous one does not that might cause this.

[Sarlock]

And why it didn't crash in the first year of the game and only in the second... and why it runs just fine locally.  If I had any coding knowledge at all, I'd be right in there helping.  Thanks for spending/wasting all of your time on this issue 

[ӔO]

is it the same under both 32bit and 64bit operation?


---
actually the game run locally on 64bit crashes in the same exact place as the server does.


---
yeah, so the when I setup a local server and client, both using 32bit, the game went into June just fine, while a 64bit client crashed as soon as June started.

Also, there is something odd about the forest on the server side, but it doesn't seem to affect the client side.

---

64bit client crashed again in July when joining a game that has already progressed beyond the point where it would have normally crashed in June.

[jamespetts]

Thank you both very much for your input, and especially AEO for the testing. May I inquire - what is odd about the forest on the server side?

Edit: Bizarrely, I have now managed to connect with my debug build and neither it nor the server have crashed. It is now June 2:01.

Edit 2: To confirm, the server is now back up. However, I seem to have noticed a pattern. The ability to connect and remain connected appears to occur after connecting to the troubled game and then disconnecting again before getting to the point where it crashes. Last night, while testing with a local server, I accidentally connected to the Bridgewater-Brunel server for a very short time, but disconnected before reaching the month end. I seem to remember something similar happening before the bug mysteriously disappeared last time. That suggests some issue with the save/load code (which might also explain why the viewport has ended up in the middle of the sea on loading rather than next to the island with the name of the server on it. Quite where this problem is, however, is a mystery to me as yet.

[asaphxiix]

Thanks James! So we have a work-around. Should we as a precaution leave and reconnect at the very end of Feb, May, Aug, and Nov?

Sadly, I don't seem to be able to change my company's name.

Also I wonder, should I build my roads over existing roads (and have them only open to players I give access to, who'll pay toll), or make new roads?

[ӔO]

Particularly, the forest is not rendered the same between server and client.
Left: 32 bit client
Middle: 64bit client
Right: 32 bit server

[AP]

Also I wonder, should I build my roads over existing roads (and have them only open to players I give access to, who'll pay toll), or make new roads?

If you are turnpiking an existing right of way I suggest you would have to give access to all players.

[asaphxiix]

If you are turnpiking an existing right of way I suggest you would have to give access to all players.

Fair enough, I won't mind, of course, except sometimes people leave the game etc.

[TurfIt]

Particularly, the forest is not rendered the same between server and client.

The positions of individual trees on a tile are not saved; Reduces the size of the save file. This will result in a different look between client/server, or even just saving/loading a game locally. The number of trees on a tile, and their ages is important for client/server sync, and so is saved.

[ӔO]

The positions of individual trees on a tile are not saved; Reduces the size of the save file. This will result in a different look between client/server, or even just saving/loading a game locally. The number of trees on a tile, and their ages is important for client/server sync, and so is saved.

ah, gotcha. It does look like the server has rows of trees with blank tiles inbetween, where as the clients have all tiles filled, but that explains it.

[AP]

It keeps crashing at the start of Aug 1751. Is there any point us persisting at the moment?

[asaphxiix]

trying the workaround now.

Edit: Alas, to no avail

[ӔO]

I think I reverted it to the beginning of July, if anyone wants to play.

interestingly, the 64bit client doesn't crash in August when connected to the server, nor locally.

---
Also, there is something odd about this new July... I don't recall connecting at the beginning of July, but the changes I've made are there.
https://dl.dropboxusercontent.com/u/17111233/client2-network_aug.sve
https://dl.dropboxusercontent.com/u/17111233/client1-network_new_july.sve
take a look at convoy (142)'s position between july-august and new july.

[Sarlock]

Kept crashing at the end of September.  I quit the game at 6:23:50, just before it flipped to October, then rejoined and the game resumed at 0:00 October and seems to now be happily progressing in to October without crashing.  I'll try again at the end of October and see what happens (assuming I'm the only player online).

[jamespetts]

End of October should not crash, as this is not a season boundary: the end of November may well crash, as this is the autumn/winter boundary. This is all very odd.

[asaphxiix]

aye, I had it at July as well. After a few joins it didn't crash. Perhaps it's an end-of-month issue.