From 6fae194ab51afad3171f6d2cdecc59649d271bc6 Mon Sep 17 00:00:00 2001
From: Augusto Barbosa Villar Silva
 <64085472+augustovillar@users.noreply.github.com>
Date: Tue, 16 Jun 2026 07:59:07 -0300
Subject: [PATCH] wcag/ChunkParser: scale Type3 glyph advances by /FontMatrix,
 not hardcoded 1/1000

PDFont.getWidth(code) returns the raw /Widths value in the font's glyph space.
For Type3 fonts glyph space is defined by /FontMatrix, not 1/1000, so the
hardcoded /1000 divisor produced advances off by (1000 * FontMatrix[0]) for any
Type3 font whose FontMatrix != 1/1000 -- corrupting TextPiece end positions and
text-chunk bounding boxes (dropped duplicate glyphs / spurious spaces downstream).

Use the horizontal /FontMatrix component for a PDType3Font (null/length/zero
guarded), 1/1000 otherwise. The TJ-array displacement path (obj.getReal()/1000,
text-space thousandths and font-independent) is left unchanged.
---
 .../gf/model/factory/chunks/ChunkParser.java      | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/wcag-validation/src/main/java/org/verapdf/gf/model/factory/chunks/ChunkParser.java b/wcag-validation/src/main/java/org/verapdf/gf/model/factory/chunks/ChunkParser.java
index e98a4f492..6937452b7 100644
--- a/wcag-validation/src/main/java/org/verapdf/gf/model/factory/chunks/ChunkParser.java
+++ b/wcag-validation/src/main/java/org/verapdf/gf/model/factory/chunks/ChunkParser.java
@@ -38,6 +38,8 @@
 import org.verapdf.pd.font.PDFontDescriptor;
 import org.verapdf.pd.images.PDXForm;
 import org.verapdf.pd.images.PDXObject;
+import org.verapdf.pd.font.PDFont;
+import org.verapdf.pd.font.type3.PDType3Font;
 import org.verapdf.pd.optionalcontent.PDOCMDDictionary;
 import org.verapdf.pd.optionalcontent.PDOptionalContentProperties;
 import org.verapdf.tools.StaticResources;
@@ -856,6 +858,7 @@ private TextPieces parseTextShowArgument(COSBase argument) {
 	private void parseString(COSString string, TextPieces textPieces) {
 		byte[] bytes = string.get();
 		try (InputStream inputStream = new ByteArrayInputStream(bytes)) {
+			double glyphSpaceToTextSpace = getGlyphWidthToTextSpaceScale(graphicsState.getTextState().getTextFont());
 			while (inputStream.available() > 0) {
 				int code = graphicsState.getTextState().getTextFont().readCode(inputStream);
 				Double width = graphicsState.getTextState().getTextFont().getWidth(code);
@@ -868,7 +871,7 @@ private void parseString(COSString string, TextPieces textPieces) {
 								graphicsState.getTextState().getWordSpacing() : 0)) *
 								graphicsState.getTextState().getHorizontalScaling();
                 width = width *
-                        graphicsState.getTextState().getTextFontSize() / 1000 *
+                        graphicsState.getTextState().getTextFontSize() * glyphSpaceToTextSpace *
                         graphicsState.getTextState().getHorizontalScaling();
                 String value = graphicsState.getTextState().getTextFont().toUnicode(code);
 				String result = value;
@@ -887,6 +890,16 @@ private void parseString(COSString string, TextPieces textPieces) {
 		}
 	}
 
+		private static double getGlyphWidthToTextSpaceScale(PDFont font) {
+		if (font instanceof PDType3Font) {
+			double[] fontMatrix = ((PDType3Font) font).getFontMatrix();
+			if (fontMatrix != null && fontMatrix.length >= 1 && fontMatrix[0] != 0.0) {
+				return fontMatrix[0];
+			}
+		}
+		return 1.0 / 1000.0;
+	}
+
 	private TextChunk createTextChunk(List<COSBase> arguments, String operatorType, int operatorIndex) {
 		org.verapdf.pd.font.PDFont font = graphicsState.getTextState().getTextFont();
 		COSBase argument = TextChunksHelper.getArgument(arguments, operatorType);