DarkRadiant 3.0.0 pre-release test

[jonri]

On 4/7/2022 at 2:57 PM, greebo said:

I have no clue yet about the lights being too dim or even missing

I think at least part of the problem might be related to ambient lights.  I built a quick little test map, and it seems like ambient lights are treated like any other point light.  Instead of providing even lighting, they are illuminating surfaces according to their normal and have a falloff.

EDIT: added https://bugs.thedarkmod.com/view.php?id=5936

Edited April 9, 2022 by jonri
add link to bug

[Frost_Salamander]

I tried the Github integration again.  When I try to 'Sync/Integrate server changes' I get this:

Git Exception: too many redirects or authentication replays

 

[greebo]

Are you on Linux? I haven't looked into authentication matters in Linux. In Windows, DarkRadiant will take the credentials stored in Windows' Password Store, so git can authenticate against the remote.

I was meaning to add some entry box to support entering the credentials, but I haven't got around to do so.

[Frost_Salamander]

1 minute ago, greebo said:

Are you on Linux? I haven't looked into authentication matters in Linux. In Windows, DarkRadiant will take the credentials stored in Windows' Password Store, so git can authenticate against the remote.

I was meaning to add some entry box to support entering the credentials, but I haven't got around to do so.

No I'm on Windows 10.

[greebo]

Usually Git for Windows will store the credentials of the remote in the windows credential store, something like git:https://github.com. If DR can find that entry, it will use the credentials from there. But maybe your git client is not storing anything in there?

Anyways, the authentication entry box needs to be implemented some day.

[Spooks]

I'd definitely appreciate it if the light diamonds weren't opaque. Back to wireframe or if translucency can be worked into it that'd be fine, too. Really annoying as they are with regards to candles and small objects like that, you can't do clutter with those things getting in your way.

Something a bit more important than that, though. 3.0.0 either has some crazy lack of optimization in the new render code or there is a memory leak somewhere in there.

I opened a heavy map (any finished map basically, my map has 14k brushes and ~5.5k entities for example). I kept the windows task manager on top to track the RAM usage and the more I moved in the 3D view the higher the "Working Set (Memory)" went. It didn't matter what render mode I was in, looking around was faster than previous DR versions but moving with the freelook caused heavy stuttering. Worse yet, zooming out the orthoview stuttered like hell and the RAM usage quickly spiked up to the max 6 GB my system has. Bam, computer freezes, have to push the reset button with a paperclip.

It's replicable, I made it happen twice. First time with the original release and now with pre2. For comparison, the copy of 2.11.0 I keep tops at 1.35 GB of memory usage on the same map and does not increase in this way at all.

[greebo]

31 minutes ago, Spooks said:

Worse yet, zooming out the orthoview stuttered like hell and the RAM usage quickly spiked up to the max 6 GB my system has. Bam, computer freezes, have to push the reset button with a paperclip.

While some lags when zooming out and bringing previously unrendered things into view might be explicable, it shouldn't eat up your RAM completely, so this sounds rather severe. I'll try to look into that.

[Spooks]

I'm still on Windows 7 per my mantis bug reports. I'm also using the portable versions to test, if that helps any. I opened Down by the Riverside's river.map ("from Project") to test out the new lighting preview mode with DR 3 earlier in the week. It also lagged like hell but mainly due to zooming out the orthoview, not so much flying around. I thought that might be normal, not my map, right, so I ignored it then. DR does die pretty quickly when I'm zooming out in King of Diamonds' kod.map and there I'd definitely know if something is out of the ordinary.

[Frost_Salamander]

4 hours ago, greebo said:

Usually Git for Windows will store the credentials of the remote in the windows credential store, something like git:https://github.com. If DR can find that entry, it will use the credentials from there. But maybe your git client is not storing anything in there?

Anyways, the authentication entry box needs to be implemented some day.

Hmm the credential seems to be there:

[greebo]

13 hours ago, Spooks said:

I'm still on Windows 7 per my mantis bug reports. I'm also using the portable versions to test, if that helps any. I opened Down by the Riverside's river.map ("from Project") to test out the new lighting preview mode with DR 3 earlier in the week. It also lagged like hell but mainly due to zooming out the orthoview, not so much flying around. I thought that might be normal, not my map, right, so I ignored it then. DR does die pretty quickly when I'm zooming out in King of Diamonds' kod.map and there I'd definitely know if something is out of the ordinary.

I downloaded Down by the Riverside and had King of Diamonds downloaded earlier, opening these maps to reproduce the problem. I tried with my recent build and with the pre2, and while DR does eat a lot of RAM (it shows 5.6 GB in Task Manager) when switching to lighting mode and shadows turned on, it stays that way and doesn't kill the process. Memory figures stay at 5.6 GB wherever I fly around in the level, regardless whether I'm in lighting mode or the regular fullbright mode.

Can somebody confirm / try doing the same? Open DR pre2 with the project setup pointing to "river" or "kingofdiamonds". Then File > Open Map from Project... and pick the kod.map or the river.map, respectively. When moving around 3D view it stutters a bit as new things come into view, RAM is increasing, but as soon as DR has "seen" everything, memory consumption should stay constant at about 5-6 GB (at least for me).

[greebo]

12 hours ago, Frost_Salamander said:

Hmm the credential seems to be there:

Can you please show me the log file when that happens? Maybe I can spot something.

[greebo]

19 hours ago, jonri said:

I think at least part of the problem might be related to ambient lights.  I built a quick little test map, and it seems like ambient lights are treated like any other point light.  Instead of providing even lighting, they are illuminating surfaces according to their normal and have a falloff.

Ambient lights are working now, this will be part of the next pre3 build.

[Frost_Salamander]

44 minutes ago, greebo said:

Can you please show me the log file when that happens? Maybe I can spot something.

Here is the console log.

I just had a thought.  I have MFA enabled on my Github account - maybe it's something to do with that? It's not an issue though with other tools that I use for this.  With either VS Code or command line I can push fine and there aren't any more prompts.

Unfortunately I can't turn it off because I use that account for work as well (hence the MFA requirement).  But maybe you could try and reproduce by enabling it briefly on your account.

main is set up to track origin/main
main is set up to track remote origin
Pushing to remote origin
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Git Exception: too many redirects or authentication replays

 

[Frost_Salamander]

I just noticed something else strange.  If you select a model in DR 3 pre2, and then try and change the skin, the model is sunk into the grid like in the screenshot, and changing the skin doesn't update the view (it just stays the same).  I tried this with a few different models and got the same result.

 

[jonri]

Tried out the latest master and the ambient lights definitely help get us closer.  It's the difference between the lighting preview being usable vs not in a lot of areas.

Separately from this, in normal fullbright mode I've had a light diamond totally disappear on me a few times, totally at random.  It's happened when changing the light color, and it disappears in both the 2D and 3D view.  I have to go hunt it down in the entity list, and when I select it, it looks like this:

Spoiler

It happened to me in pre2 as well as master.  This is without any of my translucency experiments, and other lights show up ok.  I'll write up a bug report if I can figure out how to reproduce it...

[LDAsh]

Yeah, this build does fix the "dark maps" I was talking about.
I tested again, it took ages to swap the view, but I said "wooooow...  aaaand it crashed."  But for those few moments, it looked like Doom3 in Radiant.
There were these infinitely stretched pieces of geometry flying through everything, that only appear on certain view-angles, I only ever saw that stuff with GPU glitches, like heat-death.  Something is very wrong.

[jonri]

5 hours ago, LDAsh said:

There were these infinitely stretched pieces of geometry flying through everything, that only appear on certain view-angles, I only ever saw that stuff with GPU glitches, like heat-death.  Something is very wrong.

It turns out my map has 2 func_static beams that go missing and turn into something like this down at one end of my map.  I haven't had a chance to look into it further, but I can confirm there's an issue.  Mine didn't crash though.

[OrbWeaver]

On 4/7/2022 at 7:57 PM, greebo said:

@OrbWeaver  unit tests should be working again in Linux, at least they're not immediately crashing anymore.

Confirmed. All tests are now passing on Linux.

         ~BufferObject()
         {
-            glDeleteBuffers(1, &_buffer);
+            if (_buffer != 0)
+            {
+                glDeleteBuffers(1, &_buffer);
+            }
+
             _buffer = 0;
         }

D'oh. I actually saw that _buffer was 0 in the debugger, but thought it wasn't important. Although the docs say that glDeleteBuffers should silently ignore 0, so I don't know if it's actually related to the crash.

[greebo]

On 4/11/2022 at 8:01 AM, LDAsh said:

There were these infinitely stretched pieces of geometry flying through everything, that only appear on certain view-angles, I only ever saw that stuff with GPU glitches, like heat-death.  Something is very wrong.

On 4/11/2022 at 1:34 PM, jonri said:

It turns out my map has 2 func_static beams that go missing and turn into something like this down at one end of my map.  I haven't had a chance to look into it further, but I can confirm there's an issue.  Mine didn't crash though.

I am familiar with this problem, but I haven't been able to track it down. It definitely has to be fixed before release, it's corrupting the view. I assume it's a problem in the synchronisation code that is managing the geometry data stored in the GPU buffer. At some point, things can go out of sync and those stretched geometry appears - it disappears once the map is reloaded. Everytime of the few times it happened on me, I tried to repro it but I failed, so if anybody can give me any insight on repro steps, I'd be grateful.

[Spooks]

On 4/10/2022 at 11:53 AM, greebo said:

Can somebody confirm / try doing the same? Open DR pre2 with the project setup pointing to "river" or "kingofdiamonds". Then File > Open Map from Project... and pick the kod.map or the river.map, respectively. When moving around 3D view it stutters a bit as new things come into view, RAM is increasing, but as soon as DR has "seen" everything, memory consumption should stay constant at about 5-6 GB (at least for me).

In addition to this, can anyone testing compare the RAM usage between an old version of DR and this one? If there is still something like 3 to 4 GB of discrepancy, I'd say it's still cause for concern.

[greebo]

On 4/10/2022 at 11:43 AM, Frost_Salamander said:

I just had a thought.  I have MFA enabled on my Github account - maybe it's something to do with that? It's not an issue though with other tools that I use for this.  With either VS Code or command line I can push fine and there aren't any more prompts.

I just tried with MFA enabled on my own Github account. It doesn't prompt me for anything whatsoever, I assume that's because the locally stored auth token is sufficient for pushing content.

It's difficult to see what's going on between git and the remote server without debugging or seeing the network traffic. Not sure if I can do anything here.

Do you happen to have a gitlab repository, perhaps, to see if it's working for you there?

[greebo]

A new 3.0.0pre3 build is available with a couple of fixes, see first post.

[IZaRTaX]

On 4/17/2022 at 9:52 AM, greebo said:

A new 3.0.0pre3 build is available with a couple of fixes, see first post.

Thank you Greebo for the release but now the "render in preview lighting mode" causes radiant crash I have these setting for my ambient light:

Edit: My bad I forgot to update the TDM files now it working

Edited April 18, 2022 by IZaRTaX

[Frost_Salamander]

On 4/17/2022 at 8:28 AM, greebo said:

I just tried with MFA enabled on my own Github account. It doesn't prompt me for anything whatsoever, I assume that's because the locally stored auth token is sufficient for pushing content.

It's difficult to see what's going on between git and the remote server without debugging or seeing the network traffic. Not sure if I can do anything here.

Do you happen to have a gitlab repository, perhaps, to see if it's working for you there?

I'm not really sure what is going on here either.  I don't use Gitlab so not sure of the value of trying that.  I selfishly want to get this working for my own requirements, which aren't terribly unique I wouldn't think

I tried using a different account that isn't using MFA and I get the same problem. Here are the steps I tried:

• Delete any credentials for Github in the Windows credential store
• clone the repo from Git Bash command prompt (I have Git for Windows installed).  i.e. 'git clone https://....'
• open the map in DR and make/save a change
• commit the change in DR
• try to push using command line in order to authenticate (i.e. 'git push').  This brings up a prompt where I can authenticate to Github using the browser or a token.  With either option, you need to authorise the 'Git Credential Manager' app access to your repo.
• Click all the right buttons, and the push is successful.
• make another change in DR, save it and commit in DR
• try to 'sync/integrate server changes' in DR, and you get the 'too many redirects or authentication replays' error.

Also  when doing this no new credentials ended up in the Windows credential store.  The DR console log said it found some somewhere though:

Pushing to remote origin
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Found credentials for resource git:https://github.com in the credential store
Git Exception: too many redirects or authentication replays

 

 

Edited April 23, 2022 by Frost_Salamander

[greebo]

2 hours ago, Frost_Salamander said:

I'm not really sure what is going on here either.  I don't use Gitlab so not sure of the value of trying that.  I selfishly want to get this working for my own requirements, which aren't terribly unique I wouldn't think

I have to test this on another machine some time, one that doesn't have credentials saved at all, to have any chance of seeing what's going on.

Obviously there is something saved in the credstore, DR is retrieving non-empty strings for username/password, otherwise the message wouldn't be there.