r6585 nightly linux 64 -- land sinking into the water / eathquakes?

[richardk]

r6585 linux 64.
When I started a saved game I found that some land had sunk into the water. There was a road on that square and therefore vehicles could not find a route. This has happened near water and at the edge of the map, and I think that the sunk squares might have been flat artificial slopes, but I'm not too sure. The edge of the map has lost quite a large area into the sea.

I've had this happen in both pak128.open.r1199 & pak128.Britain.

This has only started recently, sorry but I don't know exactly which version this started at, but it just happen in the latest nightly.
Any ideas what's going wrong?

Richard

[kierongreen]

Would it be possible to show a screenshot or provide a saved game?

[richardk]

Sure -- No problem. Shall I post them here? they are a bit big.

[richardk]

OK -I've copied the relevant bit of the screen shot to a png , which should now be small enough

[kierongreen]

Very strange....

[Ters]

Do you have any idea of which revision(s) of Simutrans you used before this? Even a rough estimate?

[IgorEliezer]

People from PT-Portuguese board would rejoice on seeing these screens. They asked for quakes, accidents and all kind of simcity stuff to be put in Simutrans a ton of times.

[kierongreen]

Could conceivably be incompatibility between nightly versions to do with double height code.

[Ters]

But why just those tiles? Were they something else than plain grund_t, like fundament_t? Is it even possible to fill such an area with that?

[Dwachs]

A savegame before this happened would be very helpful to find to actual cause.

[Tazze]

I'm sorry not to report that: Although I'm windows user, I had faced this disaster in about a month ago.
I post save data for pak128 Britain.
http://simutrans-germany.com/files/upload/britain_6.17-6.28.sve

my  log

6/17       I had downloaded nightly " simwingdi_2013-06-17_v112.4_r6553 " with pak britain-1208 to make road vehicle list for Pak Britain and made save data.

6/24       I had upload to "r6555" (I upload the latest version frequently..)  but world limits in this save data had be broken a bit.

7/3         I upload to "r5676" but the bug had not fixed.

7/4 or 5  I had used this save data made road vehicle list for Pak Britain. Suddenly, Problem window turned up,the ground was broken down when I looked there.

I made the list from this save data but I stoped using this.

sorry to be late to report because I checked my log and I'm carefully against bug report.

[kierongreen]

Has anyone encountered this issue when loading a release saved game as opposed to one from a nightly?

[captain crunch]

Some observations:
* It does not only happen with save games from earlier (nightly) version
* It does not only happen on shore lines: The the lowered tile in the saved game Lengerich sunken land was a train tunnel.
* It is possible to gain land by lowering the flat land tiles on the shore, see screen shots from the saved game Lengerich - Landgewinnung

Versions:
* simutrans 112.4 current trunk from github
* pak64.german 110-0c

[kierongreen]

It is possible to gain land by lowering the flat land tiles on the shore, see screen shots from the saved game Lengerich - Landgewinnung

This is intended.

[captain crunch]

Probably related, when lowering the marked tile in the attached save game, it digs a very deep shaft, which cannot get undone.

[Ters]

Probably related, when lowering the marked tile in the attached save game, it digs a very deep shaft, which cannot get undone.

This is a relatively new bug. Unfortunately, I'm not sure which revision I built the last time (only enabled WITH_REVISION now), but it did already contain some theme stuff, yet I did not get any deep shafts with that.

[z9999+]

I'm not sure this is related or not. Tested with r6646.

Can build way on water.


Cannot dig correctly.

[kierongreen]

Both bugs shown by z9999+ confirmed on 6646. Will work on later after work if someone doesn't get there first...

[captain crunch]

Perhaps introduced with this commit in simwerkz.cc line 1273?

[kierongreen]

That bit of the commit you refer to has following changes:

Code Select Expand


+  const sint8 water_hgt = welt->get_water_hgt( pos.get_2d() );
....
-      if(  !gr1->ist_wasser()  &&  new_slope == 0  &&  new_pos.z == welt->get_water_hgt( pos.get_2d() )  &&  gr1->get_typ() != grund_t::tunnelboden  ) {
+      if(  !gr1->ist_wasser()  &&  new_slope == 0  &&  new_pos.z == water_hgt  &&  gr1->get_typ() != grund_t::tunnelboden  ) {

which should be the same before...

[Ters]

The maximum height error isn't a bug, in case anyone thinks that. The bug is the overly deep shaft created just before.

[captain crunch]

That bit of the commit you refer to has following changes:

Code Select Expand


+  const sint8 water_hgt = welt->get_water_hgt( pos.get_2d() );
....
-      if(  !gr1->ist_wasser()  &&  new_slope == 0  &&  new_pos.z == welt->get_water_hgt( pos.get_2d() )  &&  gr1->get_typ() != grund_t::tunnelboden  ) {
+      if(  !gr1->ist_wasser()  &&  new_slope == 0  &&  new_pos.z == water_hgt  &&  gr1->get_typ() != grund_t::tunnelboden  ) {

which should be the same before...

Sadly, I do not know enough about C++ to be able to affirm that, but reverting the changed source line fixes the problem of creation of the erroneous deep shafts.

The bug is the overly deep shaft created just before.

Exactly. And also building land on water as z9999+ noted.

[kierongreen]

I think I've fixed both.... and it was water_hgt not being updated when water height was changed, then code after relying on old value. Committed in 6649.

[kierongreen]

Incidentally, while the bugs described by z9999+ have been fixed, can anyone confirm whether the original bugs at the top of this thread are still present?

[captain crunch]

Yes, the original bug is still not fixed:
* OK
* Rotated map twice, land sunk

[kierongreen]

Ok thanks - bug is linked to http://forum.simutrans.com/index.php?topic=12000.0 I'm trying to find out a solution that works for all tiles. Sometimes I wish that someone hadn't made the decision not to save the height grid with saved games...

[Ters]

Yes, the original bug is still not fixed:

I'm not 100 % sure this is the same as the bug reported in the first post since we have no information about how the landscape looked before. But it is certainly no impossibility. In any case, they are both related to what kierongreen links to.

[kierongreen]

Ok, I think that should be bug fixed in 6655. Can people please particularly check for any more slope errors. I think I've got them all now but you never know where more will be lurking!