ShaderCodeCursor Semicolon Truncation Corrupting Loop Statements

Start with packages/dev/core/src/Engines/Processors/shaderCodeCursor.ts, especially the semicolon-splitting logic around lines 55–72, and reproduce the issue with the for-loop examples in the report. Verify that empty fragments preserve omitted loop expressions and consecutive semicolons, while ordinary semicolon-separated lines remain unchanged.

Issue : ShaderCodeCursor Semicolon Truncation Corrupting Loop Statements

• Package: @babylonjs/core
• Subsystem: ShaderProcessor, Shader Lexical Preprocessing (GLSL & WGSL)
• Affected File: packages/dev/core/src/Engines/Processors/shaderCodeCursor.ts (Lines 55–72)
• Status: Preprocessor bug on master branch

1. Problem Description

Babylon.js includes an internal shader preprocessor that standardizes shader code before feeding it to GLSL or WGSL compilers. The line cursor splits shader lines on semicolons (;) to handle multiple statements on a single line.

However, when a shader contains standard C-style loops with omitted expressions—such as for (;;), for (int i = 0;; i++), or for (int i = 0; i < 10;), or consecutive semicolons ;;—the tokenizer discards empty fragments without preserving their semicolon delimiters.

Consequently:

• for (;;) becomes for (;) (missing the second semicolon)
• for (int i = 0;; i++) becomes for (int i = 0; i++)

Both result in invalid shader syntax, causing shader compilation failures. This routinely breaks shaders transpiled by Slang, SPIRV-Cross, or DirectXShaderCompiler (DXC).

2. Root Cause Analysis

In packages/dev/core/src/Engines/Processors/shaderCodeCursor.ts:

} else {
    // Semicolon in the middle of the line
    const split = line.split(";");

    for (let index = 0; index < split.length; index++) {
        let subLine = split[index];

        if (!subLine) {
            continue; // BUG: Drops the delimiter token for empty fragments!
        }

        subLine = subLine.trim();

        if (!subLine) {
            continue; // BUG: Drops empty whitespace-only fragments!
        }

        this._lines.push(subLine + (index !== split.length - 1 ? ";" : ""));
    }
}

Tracing for (;;):

1. split = ["for (", "", ")"]
2. index = 0: subLine = "for (" -> pushes "for (;"
3. index = 1: subLine = "" -> if (!subLine) continue; (skipped completely!)
4. index = 2: subLine = ")" -> pushes ")"
5. Recombined output: for (;) — an illegal statement in GLSL and WGSL.

3. Step-by-Step Reproduction

import { ShaderCodeCursor } from "@babylonjs/core/Engines/Processors/shaderCodeCursor";

const cursor = new ShaderCodeCursor();
cursor.lines = [
    "for (;;)",
    "for (int i = 0;; i++)"
];

// Read back processed lines
const output: string[] = [];
while (cursor.canRead) {
    cursor.lineIndex++;
    output.push(cursor.currentLine);
}

console.log("Processed output:", output);
// Actual:   ["for (;", ")", "for (int i = 0;", " i++)"]
// Expected: preserved dual semicolons suitable for loop header syntax

4. Code Fix & Patch

--- a/packages/dev/core/src/Engines/Processors/shaderCodeCursor.ts
+++ b/packages/dev/core/src/Engines/Processors/shaderCodeCursor.ts
@@ -57,17 +57,11 @@ export class ShaderCodeCursor {
                 for (let index = 0; index < split.length; index++) {
                     let subLine = split[index];
 
-                    if (!subLine) {
-                        continue;
-                    }
-
-                    subLine = subLine.trim();
-
-                    if (!subLine) {
-                        continue;
+                    if (subLine !== undefined) {
+                        subLine = subLine.trim();
                     }
 
-                    this._lines.push(subLine + (index !== split.length - 1 ? ";" : ""));
+                    this._lines.push((subLine || "") + (index !== split.length - 1 ? ";" : ""));
                 }
             }
         }

5. Contributor Value & Profile Impact

• Why it matters: Modern shader pipelines rely heavily on transpilation tools (e.g. WebGPU Slang pipelines). A bug in core line tokenization silently corrupts transpiled shaders.
• Skills demonstrated: Compiler front-end tokenization, shader preprocessing pipelines, cross-compiler compatibility.