Repository navigation
Fix GL calls WebGL rejects (black lit surfaces) + probe for the freeze - #24
Merged
Merged
Conversation
The probe showed the in-game trails are gone (frames are cleared), but a GL error (0x500, INVALID_ENUM) every frame, lit world surfaces rendering black, and a freeze after the first chamber loads. Extend it so the next diagnostics pin all three down: - Error hunt: for three short windows (frames 6-8, 200-203, 450-453) every WebGL call is followed by glGetError and the rejected calls are logged with their arguments. The wrappers are installed only for the window and removed after, so normal frames pay nothing. - Texture uploads: the first texSubImage2D/texImage2D calls are logged with size, format, UNPACK_ROW_LENGTH and a sample of their bytes (mean, share of non-zero), to see whether lightmap data reaches WebGL. - Heartbeat: an "alive" line with frame count and fps about every 10 s, so a freeze shows as the moment the lines stop. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qy5DKdwTqGQqScWvJyNcft
…robe The GL probe named the calls WebGL rejects in game: - texImage2D(GL_RGB8 internal format, GL_RGBA data) -> INVALID_OPERATION. D3DFMT_X8R8G8B8 is 4-byte BGRX data kept in a 3-channel GL_RGB8 texture, a pairing GLES 3 / WebGL 2 do not allow, so every such texture (normal maps, the engine's 1x1 white/grey/black fallback textures) had no image and sampled black. On web X8R8G8B8 is reported unsupported, so the material system converts those images to R8G8B8 when it loads them; convert_texture also promotes RGB8 to RGBA8 whenever 4-channel data arrives, so no texture is left without an image. - texParameter(GL_TEXTURE_BORDER_COLOR), the GL_CLAMP_TO_BORDER wrap mode and enable/disable(GL_FRAMEBUFFER_SRGB) -> INVALID_ENUM every frame. WebGL has none of them: skip the border colour, clamp to edge instead, and leave the sRGB write to the shader (as the web build already does). Freeze: frames stopped about 12 s into the first chamber. The probe now logs frames slower than 1.5 s with their draws, uploads and GPU buffer read-backs, counts read-backs in the heartbeat, adds the framebuffer status to INVALID_FRAMEBUFFER_OPERATION errors, and the page logs when the render thread stops presenting frames and when it resumes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qy5DKdwTqGQqScWvJyNcft
matthewcodergamer
changed the title
…ndows The probe showed clears and draws failing with framebuffer status 0x8CD9 (FRAMEBUFFER_INCOMPLETE_DIMENSIONS): WebGL 2 still requires every attachment of a framebuffer to have the same size, while Source renders into smaller targets (1024x512, 292x135, ...) with the full-screen 1168x540 depth buffer bound, as D3D9 allows. Everything drawn into those framebuffers was lost. On web such a framebuffer now gets a private depth buffer of the colour target's size (created once per size); those passes only touch the top-left of the big depth buffer and clear depth anyway. The menu also ran at 6-12 s per frame in the last run. Add profile windows (frames 11-13, 60-63, 240-243, or up to a slow frame) that count WebGL calls by name and time spent inside them versus the whole frame, so the next diagnostics show whether the time goes to WebGL or to the engine. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qy5DKdwTqGQqScWvJyNcft
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Black lit surfaces
The GL probe named the calls WebGL rejects in game:
texImage2D(GL_RGB8, …, GL_RGBA)→ INVALID_OPERATION.D3DFMT_X8R8G8B8is 4-byte BGRX data stored in a 3-channelGL_RGB8texture. GLES 3 / WebGL 2 don't allow that pairing, so every such texture had no image and sampled black. That covers normal maps and the engine's 1×1 white/grey/black fallback textures.X8R8G8B8is now reported unsupported, so the material system converts those images toR8G8B8when it loads them.convert_texturealso promotesRGB8toRGBA8whenever 4-channel data arrives, so no texture is left without an image.GL_TEXTURE_BORDER_COLOR, theGL_CLAMP_TO_BORDERwrap mode andGL_FRAMEBUFFER_SRGB→ INVALID_ENUM every frame. WebGL has none of these. On web the border colour is skipped, wrapping clamps to edge instead, and the sRGB write stays in the shader (the web build already does this).Probe for the freeze
Frames stopped about 12 s into the first chamber. The probe now logs:
The first commit added the original probe: an error hunt, texture-upload samples and the heartbeat.
Checks
em++ -fsyntax-only(Emscripten 4.0.9) passes oncglmtex.cpp,dxabstract.cppandglmgr.cpp.node --check emscripten/pre.jspasses.🤖 Generated with Claude Code
https://claude.ai/code/session_01Qy5DKdwTqGQqScWvJyNcft
Generated by Claude Code