zoot

That's my point, lol. Who knows where features like units on walls and triggers would have been now if everyone outside the team had not been precluded from those discussions?

No, but I won't bother working on it if there isn't really interest.

Someone reported a bug with ARB shaders here.

There's no new commits on it?

Yeah, that makes some sense. Still, there is something to be said for solving problems (such as formalizing an idea to the point where it's worth filing a ticket for) in a larger forum. The fact that I can't find any prior discussion on the public forum of "siege mines" or "sap points" tells me there is a lack of infrastructure/tradition for doing this.It's not a big deal for the individual idea, but if you imagine a team of people doing it consistently for a period of years, the gap in understanding grows so large that those on the "outside" are resigned to sit back and watch and expect that someone else have everything covered. Probably the easiest way to do it would be to place a decal every x meters (or whatever the in-game unit of distance is ) from the sap point as the construction of the tunnel progressed. In the interest of performance, I guess they should "fade away" after the tunnel has been destroyed.

I agree, it would look nicer, but it would prevent you from building buildings very closely together (i.e. no building snapping).

Yeah, I thought the effect became rather indiscernible at a distance. (Maybe anisotropic filtering would help, but it's a performance hog.) I think the object has to be in view for the crash to happen, but it is quite likely that other objects are affected too.

(Upped the intensity a slight bit.)

Okay, it happens specifically when trying to place a Celtic "mill" (dropsite for wood etc.) It also crashes when trying to place it (structures/celt/mill.xml) in Atlas.

I got a similar error on Acropolis 1 a little while into the game: ShaderManager.cpp(474): Assertion failed: "0 && (L"Can't find a usable technique")" Assertion failed: "0 && (L"Can't find a usable technique")" Location: ShaderManager.cpp:474 (NewEffect) Call stack: (0x82223a) ./pyrogenesis() [0x82223a] (0x7c78b1) ./pyrogenesis() [0x7c78b1] (0x7c858e) ./pyrogenesis() [0x7c858e] (0x7c8863) ./pyrogenesis() [0x7c8863] (0x6167ec) ./pyrogenesis() [0x6167ec] (0x617f8d) ./pyrogenesis() [0x617f8d] (0x6ec78a) ./pyrogenesis() [0x6ec78a] (0x6c83c1) ./pyrogenesis() [0x6c83c1] (0x69989b) ./pyrogenesis() [0x69989b] (0x69cc82) ./pyrogenesis() [0x69cc82] (0x69d61f) ./pyrogenesis() [0x69d61f] (0x596c81) ./pyrogenesis() [0x596c81] (0x41df63) ./pyrogenesis() [0x41df63] (0x413a47) ./pyrogenesis() [0x413a47] (0x7f7d1417c76d) /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7f7d1417c76d] (0x41c575) ./pyrogenesis() [0x41c575] errno = 0 (Try again later) OS error = ? (This was with GLSL and everything on.)

Do commit! Want to check out that rock

Can you zoom in on it without crashing? I had to zoom make it crash, probably due to the level-of-detail feature.

Yep, he committed them yesterday.

I got this error when trying to open Gen.Kenobi's rock in Atlas: ShaderProgram.cpp(901): Assertion failed: "(m_StreamFlags & ~m_ValidStreams) == 0" Assertion failed: "(m_StreamFlags & ~m_ValidStreams) == 0" Location: ShaderProgram.cpp:901 (AssertPointersBound) Call stack: (0x82223a) ./pyrogenesis() [0x82223a] (0x7c78b1) ./pyrogenesis() [0x7c78b1] (0x7c858e) ./pyrogenesis() [0x7c858e] (0x7c8863) ./pyrogenesis() [0x7c8863] (0x666127) ./pyrogenesis() [0x666127] (0x68d1dc) ./pyrogenesis() [0x68d1dc] (0x699b56) ./pyrogenesis() [0x699b56] (0x69ccac) ./pyrogenesis() [0x69ccac] (0x69d61f) ./pyrogenesis() [0x69d61f] (0x596c81) ./pyrogenesis() [0x596c81] (0x6f3df8) ./pyrogenesis() [0x6f3df8] (0x6efeff) ./pyrogenesis() [0x6efeff] (0x7f5a8986ee9a) /lib/x86_64-linux-gnu/libpthread.so.0(+0x7e9a) [0x7f5a8986ee9a] (0x7f5a8959c4bd) /lib/x86_64-linux-gnu/libc.so.6(clone+0x6d) [0x7f5a8959c4bd] errno = 0 (No error reported here) OS error = ?

I must admit I don't see how turning people away or not bothering reaching out to them makes priority items any more likely to be worked on, but it isn't my call.

IMO, there is an unfortunate tendency to think that if no one on the team has time, it can't be done. That is not an effective utilization of open source "mojo". May I suggest creating a "development blog" on the new website, whose purpose would be to present problems like this in an enticing way to potential contributors in the blogosphere? I might volunteer to write for such a thing and aggregators like this one can help propagate it. I can't guarantee it will work, but it seems it would be worth a shot.

Tiling artifacts are apparent on big stretches of terrain using the same texture. This, IMO, is sub-par. One possible solution is to have several different textures for the same type of terrain (a bit like actor 'variants'). Programmatically, this should be easy to do with myconid's terrain XML. However, I think it would be a lot trouble for someone to go through all the terrain types and create variants for all of them, given that each variant needs to fit seamlessly with the others. A different solution might involve texture synthesis (see below). This would be substantially harder to implement and I have no idea what the performance impact would be, but it would make it easy for everyone (including modders) to get nice-looking, tile-free terrains. Other ideas/comments?

Don't think it would be that hard for someone who wants to do it. The trickiest stuff can probably be reused from the wall implementation. The main remaining piece of logic is then the one for making the "creeping" texture.

The images? The ones above were taken from its low-res 'hazcam' which is used by its on-board computer to avoid unforeseen hazards like pits and rocks. Once its mast is deployed, we'll get hi-res true color images (heck, they'll even be stereoscopic).

Conditional defines now on review. It adds a new <conditional_define> element to material XML, like this. It also instantiates a new object in the CMaterial class for holding conditional defines here. It can be accessed through the GetConditionalDefines() method here and new conditional defines can be added to the object with the AddConditionalDefine() method here. The new element is parsed here. The element takes a number of attributes, one of which is 'type'. The 'type' attribute specifies, in effect, a comparison operator for the conditional define. Currently, only one such operator 'type' is recognized, namely 'draw_range'. The 'draw_range' operator loads two config values (a 'min' and a 'max' value) specified in the 'conf' attribute of the 'conditional_define' element. If no 'conf' attribute is present, it instead tries loading the 'min' and 'max' value from the element attributes. If the 'min' and 'max' values are greater than or equal to zero, they are added to the material as a 'define' for shaders. When the 'type' attribute has been processed, the conditional define is then added to the material. While rendering, the engine iterates through the conditional defines here. If a given conditional define is of the 'draw_range' type, the engine calculates the distance from the model being drawn to the camera. If the distance is within the 'min' and 'max' ranges specified for the conditional define, the 'value' attribute of the conditional define is added as a define with the name given by the 'name' attribute. I didn't come across any glaring errors in any of the above.

What does it entail? If it's just using the mining animation when a specific unit attacks a specific structure, it can't be too hard.

Would it be messy to make the conditional defines work for both cases?

I see. But doesn't conditional defines then make the quality level system obsolete? Can quality levels do anything that conditional defines cannot?

I must admit I'm a bit ambivalent about the whole notion of trying to map quality to an absolute, global scale. What if you have a well-performing effect called super mapping that looks excellent (quality=10) and a separate low-performing effect called meh mapping that doesn't look too good (quality=5) - if you want to improve performance of meh mapping, you can try and reduce the quality config setting (so as to select a better performing alternative), but then you also risk reducing the quality of super mapping, which isn't really necessary because it performs well already?

Please see here.