Skip to content

3.5-alpha16 (5th December 2018)

Pre-release
Pre-release
Compare
Choose a tag to compare
@meag meag released this 05 Dec 20:08
· 1358 commits to meag/renderer since this release

READ FIRST

Major rendering engine update, expect lots of bugs, DO NOT USE for important games. It has had very limited testing and isn't ready yet.

Source-code if you want to build on other systems is available here, process to build is:

git clone https://github.com/meag/ezquake-source new-renderer
cd new-renderer
git checkout meag/renderer

make clean && CLASSIC_OPENGL_ONLY=1 make -j3
make clean && MODERN_OPENGL_ONLY=1 make -j3

Changes in 3.5-alpha16

  • ciscon: "if you use gl_gammacorrection 1, you have to do a vid_restart after startup. or i should say, if you do, it gets reset"
  • ciscon: "r_fullbrightskins is either on or off it appears, anything below 1 is the same"
  • ciscon/dev: alias model textures corrupt/incorrect
  • ginzberg: "is it known that gl_detail 1 is sad in 3.5?"
  • meag: /r_drawflat affected brush models
  • meag: /cl_fakeshaft disabled during intermission
  • meag: added /r_fx_geometry for outlining world based on surface normals, not surfaces. requires /vid_framebuffer 1
  • meag: removed cached memory in hunk (skins & models)
  • meag: disabled logging console during match in competitive rulesets
  • meag: /cl_bufferwait re-enabled
  • meag: /r_tracker_inconsole messages now have the image in the console
  • meag: legacy variables (renamed) now registered earlier, so old values should set new console variables.
  • meag: tracker is now also a hud element

Known/reported issues

Open issues/reports from previous versions

  • dev: screen flickering. (... cannot replicate)
    • initially thought to be only on linux, but seems it happens on windows too after long uptime.
    • (dev reports that mur has same issue)
    • (we now think this is a hud-rendering issue but haven't got to the bottom of it)
  • dev: crash when issuing /vid_restart while fast-forwarding a demo. Error message looks like it might be freeing() memory twice. (command was gl_externalTextures_world 1; gl_max_size 32768; gl_texturemode GL_LINEAR_MIPMAP_LINEAR; gl_loadlitfiles 1; gl_outline 1; vid_restart)
  • lordlame: complaints about performance on radeon 6990 (reports alpha3 runs better)

Latest version

  • bgnr: fov distortion on vid_restart
  • bgnr: gamma doesn't work, reports "gamma size broken, gamma wont work" (likely to be linux specific)
  • confy: show speed is broken, show speed2 works but takes over whole screen (cannot replicate, waiting on config)
  • meag: /vid_framebuffer & /vid_gamma_correction is double-applying the gamma correction
  • meag: /r_lerpframes & md3 models... double-check
  • meag: core profile rendering broken on AMD (narrowed it down to draw_world)
  • meag: stop switching programs when drawing different hud elements (especially circle crap)

Changes in upcoming version

  • includes all changes up to and including v3.1
  • bgnr: reports that gfx.wad consolefont is not loaded on startup
  • bps: "bug on alpha15 where it glows on the ground on couple of places, like the glow from a quad, currently only on ztndm3" - have downloaded the .mvd, waiting on further details
  • ciscon: reports that nailtrails disappear when using simple item textures of higher dimensions (has supplied .zip of new simple items, confirmed that removing them fixes problem)
  • ciscon: gl_outline & gl_multisamples conflict in glsl - gl_outline_width was not being applied in modern
  • dopeskillz: "I can do about 12 /vid_restarts and then ezQuake quits with the message 'Too many graphics buffers allocated'"
  • drzel: /scr_newhud_clear should be forced on (defaults to 0 because of software according to original commit)
  • meag: /gl_simpleitems 0, /r_drawflat 1 caused item brushmodels to disappear
  • meag: toggling /r_drawflat led to lightmaps on surfaces being processed twice/frame (fps drop)
  • meag: fixed potential NULL reference error when deleting VBO buffer (seemed to be classic only)
  • meag: /cl_bufferwait 1 & timedemo is broken
  • meag: powerupshell didn't match underlying model's lerp values
  • mushi: "when moving feels like 5fps" (have asked to try /cl_bufferwait 0) - has now reported that he can't replicate this - have removed /cl_bufferwait again, caused these kind of issues on AMD card
  • skydome rendering in glsl optimised
  • skydome/box rendering can use glsl in classic renderer, enabled by default (/gl_program_sky 1)
  • turb surface rendering can use glsl in classic renderer, enabled by default (/gl_program_turbsurfaces 1)
  • aliasmodels can use glsl in classic renderer, enabled by default (/gl_program_aliasmodels 1)
  • world can use glsl in classic renderer, enabled by default (/gl_program_world 1)
  • hud rendering can use glsl in classic renderer, enabled by default (/gl_program_hud 1) - incomplete, still circles etc to do
  • added /r_lightmap_lateupload to upload changes to lightmaps as they are bound (anyone who had better performance with -nomtex, this might help performance)
  • added /r_lightmap_packbytexture to change the order surface lightmaps are added to atlases (naive way to reduce number of texture switches when rendering world)
  • renamed /scr_printspeed to /scr_centerspeed
  • removed functionality: turb ripple (was demos only)
  • removed functionality: scene blur
  • removed /r_glstats, values now in hud element framestats
  • fix hunk-leak when loading or starting a new single player game
  • added hud elements for centerprint & qtv_buffer
  • classic renderer uses texture arrays for lightmaps, if available
  • don't check for updated lightmaps on non-world brushmodels
  • added /r_drawhud option, for timedemo testing (also useful for level screenshots)

Fixed in 3.5-menus

  • dirtbox: rendering of server browser's server popup window was broken (was connected to server and running server browser and having server-browser window open)
    • full console was drawn in background then painted over before server-browser
    • spaces were still drawn when drawing red text, rather than skipped as normal
    • dialog boxes were rendered as individual blocks rather than repeating tiles (very old issue)
  • meag: crosshairimage is set to an image that is not available, will not load (memory in pool not allocated)
  • meag: proportional font loading is broken (uses new spacing for spaces but still uses fixed-width font)
  • meag: server browser threading model significantly changed
  • meag: user interface functionality significantly changed

Old bugs

  • timon: coronas of rockets in water are visible when out of the water
  • meag: cheap shadows are affected by rotation of item (most noticeable on gibs)

Bugs reported, not being actioned

  • raket: http://upload.foppa.dk/files/ez35love.mkv, rendering to one x server and capturing from another on nvidia. Works on 3.1, 3.5-classic crashes immediately, 3.5-glsl gives the video attached. also breaks cs:go. no bug if ffmpeg capturing setup isn't enabled, raket suspects bug in linux nvidia drivers
  • raket: requested change: detect compositor on x-windows

Hopefully fixed (waiting for feedback)

  • bgnr: text font appears as solid colour (reports that running without config works fine, has supplied config - cannot replicate)
  • bps: crash on last seconds of dm2 (hoping this was the particle-expiring issue fixed in alpha6)
  • dimman: crash when running timedemo demo1 (...cannot replicate, believe might be same as bps)
  • ciscon: on mesa, /vid_gl_core_profile to get > 3.0 on intel card supporting 4.5 no longer works
  • dirtbox: related out of memory error when trying to connect to server (can't get this to happen, but fixed some memory leaks and memory usage should now be much lower)
  • lordlame: vid_renderer 0 > 1 > 0 loses performance increase (performance stays at '1' levels) - AMD system
  • up2: incorrect luma textures used on start.bsp (fullbright player skins layers used instead of brush model luma textures) (... could not replicate but alpha16 fixed bug in GL_BindTextures() call)

Struggling to replicate

  • blah/ciscon: increased memory usage over time (... cannot replicate)
  • nigve: 'q' in tracker isn't blue anymore (can't replicate)