Observations on 0.48 RC1

General discussion regarding the OpenMW project.
For technical support, please use the Support subforum.
ghales
Posts: 41
Joined: 25 Jul 2014, 00:25

Observations on 0.48 RC1

Post by ghales »

For an RC this is an amazingly solid release. I've had no crashes or glitches and the framerate has been a steady 60 FPS (limited), even with extended view distance and shadows enabled. This is on my midrange I5 and RX 560. There are a few minor issues, though.

Every once in a while there are white speckles along the terrain seams:
Seams.jpg
As you change position or look around, the speckles dance up and down the seams. It does not happen often, and happens in random places. If you leave, you can come back to the same area and the sparkles will be gone. I can't reproduce it deliberately, there's nothing in the log, and of course it's gone if you reload a save. :?

Another visual glitch happens fairly consistently along the edges of water:
WatersEdge.jpg
The bright spots move with the ripples in the water. I know there is work going into a better water shader, so this may be moot.

I notice that Netchs now shoot at you, rather than over your head. :D But, other attackers are still happy to shoot straight into an obstacle you are standing behind:
MissedShot.jpg
Is there a way for NPCs/creatures to check for a clear shot, before they fire off all their ammo into the terrain?

I also note that there are shadows at night which look very odd since there is no light source:
ShadowsAtNight.jpg
Finally, I found a corner case for the nav system. The skeleton outside of Omaren (NE of Sadrith Mora - Skull Crusher quest) spawns too close to the rock and gets stuck there. It justs runs in place, trying to reach you:
Omaren.jpg
Overall, this looks like a great release. Thank you all.
User avatar
psi29a
Posts: 5361
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: Observations on 0.48 RC1

Post by psi29a »

Are any of these things issues in 0.47?

Also, are they reproducible in our current 0.49 dev dailies?
User avatar
akortunov
Posts: 899
Joined: 13 Mar 2017, 13:49
Location: Samara, Russian Federation

Re: Observations on 0.48 RC1

Post by akortunov »

psi29a wrote: 03 Sep 2022, 09:10 Are any of these things issues in 0.47?
An issue with terrain seems to be a lack of position precision, I encountered such issue with OSG compiled with single precision, but did not encounter it with double precision. Did we update OSG in our dependencies?
An issue with water shader seems to exist since water shader shader was implemented. There were fixes to mitigate this issue, but it was not completely fixed. Not a regression.
An issue with shadows is an intended behaviour - IIRC, during night OpenMW's shadows system uses a fake light source. Not a regression.
An issue with skeleton seems to be not related to navmeshes at all. More likely a spawn point is just inside of rock, so spawned creature can not move because Bullet does not support one-side collisions. Not a regression.
Can not reproduce the issue with aiming, so can not tell anything about it.
ghales
Posts: 41
Joined: 25 Jul 2014, 00:25

Re: Observations on 0.48 RC1

Post by ghales »

IIRC there used to be generic Linux builds of the "nightlies" available from the Downloads page, but they aren't there now. My attempts to compile from source have not gone well, so without those binaries I really can't test the current development build.

The speckled seam issue was in 0.47, but much more common than in RC1, at least, as far as I can recall. By the way, I have a save from last night that has the sparkly seams. It's 2M zipped, but I can open a bug report and upload it if it will help.

Shadows at night is an intended behaviour???

Skeleton issue is definitely not a regression, since it was the same in 0.47. :lol: The problem does seem like a content issue, that could be fixed by a mod.

The aiming problem is very common, and easily reproduced. Just stand at the top of a set of stairs, or on a hill. I only raised the issue because fixing Netches made me happy. But, that is just the first step in getting mobs to aim properly, i.e. with ray-casting to target center of visible mass.
User avatar
akortunov
Posts: 899
Joined: 13 Mar 2017, 13:49
Location: Samara, Russian Federation

Re: Observations on 0.48 RC1

Post by akortunov »

ghales wrote: 03 Sep 2022, 16:14 Shadows at night is an intended behaviour???
Yes, it is. As I said, shadows at night use a fake light source (as in interiors) rather than completely disappear.
User avatar
AnyOldName3
Posts: 2674
Joined: 26 Nov 2015, 03:25

Re: Observations on 0.48 RC1

Post by AnyOldName3 »

Everything uses the fake light source, not just shadows, and we think we've got a pretty accurate reproduction of the path it takes at night in the original engine. In real life, the outside is too dark to see at night without street lighting, so games tend to add a fake light so you can still play them. The shadows block the fake light, though. You can make them less pronounced by adjusting weather settings so there's less diffuse light and more ambient, and once we've added soft shadows and the ability for weather to control them, it'll become possible to make shadows softer at night so they're more realistic.
ghales
Posts: 41
Joined: 25 Jul 2014, 00:25

Re: Observations on 0.48 RC1

Post by ghales »

once we've added soft shadows and the ability for weather to control them, it'll become possible to make shadows softer at night so they're more realistic.
Sounds good.
LoneWolf
Posts: 138
Joined: 26 Sep 2017, 19:13

Re: Observations on 0.48 RC1

Post by LoneWolf »

ghales wrote: 03 Sep 2022, 16:14 IIRC there used to be generic Linux builds of the "nightlies" available from the Downloads page, but they aren't there now. My attempts to compile from source have not gone well, so without those binaries I really can't test the current development build.
For ubuntu there's a ppa with latest build, for archlinux there's an aur package.

I don't know where to find generic linux dev binaries, but there's a script to build a generic image usable through docker at
https://gitlab.com/hristoast/openmw-gen ... requests/1
K1ll
Posts: 184
Joined: 06 Aug 2011, 21:54

Re: Observations on 0.48 RC1

Post by K1ll »

LoneWolf wrote: 04 Sep 2022, 17:44
ghales wrote: 03 Sep 2022, 16:14 IIRC there used to be generic Linux builds of the "nightlies" available from the Downloads page, but they aren't there now. My attempts to compile from source have not gone well, so without those binaries I really can't test the current development build.
For ubuntu there's a ppa with latest build, for archlinux there's an aur package.

I don't know where to find generic linux dev binaries, but there's a script to build a generic image usable through docker at
https://gitlab.com/hristoast/openmw-gen ... requests/1
The nightlies still exist and are available here. Not sure why the link was removed from the Downloads page.
akortunov wrote: 03 Sep 2022, 12:30
psi29a wrote: 03 Sep 2022, 09:10 Are any of these things issues in 0.47?
An issue with terrain seems to be a lack of position precision, I encountered such issue with OSG compiled with single precision, but did not encounter it with double precision. Did we update OSG in our dependencies?
...
How can OSG be build with double precision? The library shipped in the targz package is built from the "3.6" branch of https://github.com/OpenMW/osg.git with just the CMake flags -DBUILD_OSG_APPLICATIONS="OFF" -DCMAKE_BUILD_TYPE=Release .
User avatar
akortunov
Posts: 899
Joined: 13 Mar 2017, 13:49
Location: Samara, Russian Federation

Re: Observations on 0.48 RC1

Post by akortunov »

K1ll wrote: 04 Sep 2022, 20:20 How can OSG be build with double precision?
It is supposed to be by default, but our OSG's fork had a commit which forced OSG to use floats instead of doubles in matrices to increase FPS a bit. We dropped this commit some time ago.
Post Reply