stgatilov

It is much easier to assume that we have single license for assets and all the contributors agree to distribute under it when they donate their stuff to the project. Otherwise you have to track metainformation for thousands of files, and surely there will be some errors made during tracking. If I don't care about the "TDM-libre" story, why should I bother?

Maybe start with finding who are these "we" people who want to change the license to assets?... As a programmer, I definitely do not want to maintain any additional packages. And I do not want to get entangled into any kind of licensing questions. Just that I understand: A license that forbids commercial usage is considered not free (CC-NC). A license that allows commercial usage but efficiently forbids making money from it is considered free (GPL and AGPL). Yes, this is a great distinction Now someone should come and say "but hey, you can sell your support for GPL product!" UPDATE: Well, I think there is also an approach when company provides reduced version under GPL and expanded version under commercial license (like Qt). As long as reduced version is reduced enough, it seems to work fine.

A colleague of mine believes the true mirrored rendering is a deprecated feature today. Given all the problems (try to put a model which is both partly in front of and partly behind the mirror) and the fact that reflection probes are unversal, perhaps he is right.

Yes, I considered this idea. But it won't help much, I believe. If several mirrors are allowed, you can split your surface into individual planar polygons and hope for the best. But if you have 10+ polygons, then I suppose the best won't happen Internally, the material has "mirrorRenderMap" keyword which triggers rendering of subview. This subview is than copied into a texture. In the main view, the surfaces with mirror material should be rendered as any other surfaces, it can just use this dynamically generated surface. I'm not sure what exactly you want. If you want the mirrored view to be distorted by normal map, then yes: this is definitely not supported. The reflected texture is generated as if reflected by a perfect plane. Maybe we can add some kind of warp shader that deforms the image mathematically according to the normalmap, but this would be a new feature.

Nowhere. I don't think it exists. I believe the engine source code and shaders are assumed to be under GPL (except for third-party libraries, which have their license but are essentially distributed in TDM under GPL terms). All the rest is assumed to be CC-by-nc-sa.

First some people come saying the license of TDM is too free and blocks using TDM in commercial projects. Then other people come and say the license is not free enough (seriously, I wonder why). Anyway, the license will not change.

You can't imagine how it feels to me I am testing the old missions that I played many times (like Saint Lucia and New Job), and I learn so much new! I believe those people never said this stuff before Maybe it was some hidden bug which prevented them from saying all those interesting things before, and now all of a sudden it is unlocked?! I underestimated how hard it is for a non-native speaker to understand stylized English speech.

beta212-04 is released It does not contain much changes, except for the huge pack of new subtitles

reloadDecls

I tested editing inline subtitle in tdm_stlucia.subs. That's a core file and I test on SVN, so all core assets are unpacked. It works. I edited it before triggering the sound, and then triggering it showed me the new subtitle text. I also checked in the code. Whenever some .subs file is detected as modified, all the subtitles are reloaded completely. What setup do you have specifically?

Can't understand why GPU behavior depends on monitor. Maybe test all the configurations in Windowed mode with exactly the same resolution? Perhaps with Borderless/Fullscreen you get different resolution and it affects how much VRAM you have. In fact, WDDM also reserves some VRAM for itself in order to render Windows desktop. It means that there is different amount of VRAM left for TDM (which usually does not matter unless you are very low). Maybe try to set Windows display resolution the same and check if two monitors behave the same? Also I agree that image_downsize is worth a try: it should reduce VRAM demands of TDM. Same for lowering "Render Scale" in the graphics menu and disabling antialiasing. You can also try some software which tracks amount of VRAM used, maybe you get some insights from it.

I don't think there is any purpose. Windows and Linux have entirely different window/OS integration. They are separate subdirectories with completely different code So there is no wonder that such things are different between Windows and Linux. I think @duzenko spent some time improving the behavior of Windows build. On the other hand, @cabalistic switched Linux build to GLFW. I am afraid of touching this myself because 1) I'm not a Linux user and don't know much about the myriad of desktops and APIs to integrate with them, and 2) I don't want to install Linux natively and this cursor stuff is a bit special inside VM. Maybe it is a simple flag which can be set to GLFW, I don't know.

Because internally these two actions are called _jump and _mantle, and you are looking at .cfg file where internal names are used. If someone makes a submenu for configuring controllers, I guess it can be named as jump/mantle there. You probably find answer here: https://github.com/fholger/thedarkmodvr/wiki/Gamepad-support No idea what you mean exactly, so I suppose it will not. In general, experimental controller support was initially done by @cabalistic and there was no user-friendly configuration. Nothing changed since then.

How do you determine what is decl file and what is not? Actually, xdata is decl files too. You can equally argue that materials and sound shaders are not decl files and deserve a separate reload command. I guess ID guys did not want dozen of separate commands for different kind of decls.

The "light_1" has hide_distance 100, so disabling it when player is farther than 100 units away is desired behavior. I think before 2.10 the issue was not happening because back then "hide_distance" did not work if you did not specify "dist_check_period" also. In 2.10 requirement for explicit "dist_check_period" was lifted, and this LOD started working. It is indeed bad, if it is not desired behavior, then why such low hide_distance was set?

The graphical indicators is something for an artist to invent. The lightgem graphics has not been touched for ages, but maybe there is a way to add something without annoying everyone.

I did something for it. I think it should be reloadDecls... Rather logical since subtitles are decl files.

Yes, this issue definitely won't be forgotten.

You need to parse .dat files of the fonts to see the rectangles used by different characters. Then you can take .dds file of the font and somehow overlay rectangles over the image. I'm not even sure which kind of program is suitable for it.

It should work I think.

Thank you! I have linked this to 5282. Yes, I think doing this at the lower level of idDeviceContext (where virtual resolution scaling happens) is the better approach than trying to do it on the idWindow level. It seems that you assume current UIs were made for 16:9. I'm afraid the sad truth is that they were partly made for 16:9 (new ones) and partly made for 4:3 (old ones). And the old ones look stretched. I'm not even sure which kind of intended aspect ratio is more popular now. I guess I'll try to make the scaling configurable depending on settings in the root window (aka "Desktop") of the UI. Moreover, we already have some overlay images prepared for several aspect ratios (ultrawide not included). Ideally, I'd like to drop all the duplicates and have only one image with proper scaling/padding. A long story obviously. Yeah, the special handling of full-screen windows feels hacky. I guess I'll need to avoid it somehow.

@Geep Regarding 0.24 vs 0.25 font scale. I think there was purpose, though I can't say now which exactly. Maybe it was to ensure that all the text fits, maybe it was to make vertical padding symmetric. Anyway, why do you think 0.24 and 0.25 differ? I did not look in the code yet, but I think what engine does is 1) choose most suitable image, 2) render quad texture with it. Indeed, we can avoid some blurring if the pixel size of the quad exactly matches the image size. But I guess the engine counts font size in 640x480 virtual resolution. So even if the font size would be "12", it would later be stretched over whatever width you really have (if your aspect ratio is 16:9 then height would be stretched equally, but otherwise it would be stretched with different factor).

Any warning during dmap? Does your patch have [0..1] non-overlapping texcoords?

It looks like sooner or later we'll have to add settings menu for subtitles. Verbosity can be moved to it. Then there can be this big switch between background and outlined text. And a bunch of color settings for various elements. The only thing I wish to not see there is any settings which can affect text layout

I also hope the big pile of movement changes is finally over All these changes seem minor but cause heavy reaction. First we got overly lengthy arguments about it internally, now the same continues publicly. I'm not sure why everyone thinks parkour was the main motivation. In fact, I watched parkour videos to understand how the hell it should be possible to do the wall climb up as fast as @Daft Mugi wants it. Now I play Prey and its climb up is insanely fast, totally unrealistic... I guess in the end most people are annoyed by having to wait for the same boring mantling animation and prefer to regain control faster. By the way, I believe 50% of reception to such changes boils down to plain habit. When I took the first build of faster mantle it seemed crazily fast. But after we tweaked it (I don't think we made it much slower in the end) the reception gradually changed and I did not feel it that absurdly fast anymore. That's why I even suggested accelerating mantle gradually (e.g. 100 ms per release) to avoid much negative feedback