From 0a03e3d1cf557541a7937f8603821a868c5302fe Mon Sep 17 00:00:00 2001
From: Guillaume Ayoub <guillaume@courtbouillon.org>
Date: Mon, 29 May 2023 14:57:14 +0200
Subject: [PATCH] Use UTF8 indices instead of unicode indices for line split

Related to ietf-tools/datatracker#5507
---
 tests/layout/test_inline.py    | 28 ++++++++++++++++++++++++++++
 weasyprint/layout/preferred.py | 10 +++++-----
 2 files changed, 33 insertions(+), 5 deletions(-)

diff --git a/tests/layout/test_inline.py b/tests/layout/test_inline.py
index 5c5659d15..49b0eddaa 100644
--- a/tests/layout/test_inline.py
+++ b/tests/layout/test_inline.py
@@ -464,6 +464,34 @@ def test_breaking_linebox_regression_14():
     assert line2.children[1].children[0].text == 'c'
 
 
+@assert_no_logs
+def test_breaking_linebox_regression_15():
+    # Regression test for https://github.com/ietf-tools/datatracker/issues/5507
+    page, = render_pages(
+        '<style>'
+        '  @font-face {src: url(weasyprint.otf); font-family: weasyprint}'
+        '  body {font-family: weasyprint; font-size: 4px}'
+        '  pre {float: left}'
+        '</style>'
+        '<pre>ab©\n'
+        'déf\n'
+        'ghïj\n'
+        'klm</pre>')
+    html, = page.children
+    body, = html.children
+    pre, = body.children
+    line1, line2, line3, line4 = pre.children
+    assert line1.children[0].text == 'ab©'
+    assert line2.children[0].text == 'déf'
+    assert line3.children[0].text == 'ghïj'
+    assert line4.children[0].text == 'klm'
+    assert line1.children[0].width == 4 * 3
+    assert line2.children[0].width == 4 * 3
+    assert line3.children[0].width == 4 * 4
+    assert line4.children[0].width == 4 * 3
+    assert pre.width == 4 * 4
+
+
 @assert_no_logs
 def test_linebox_text():
     page, = render_pages('''
diff --git a/weasyprint/layout/preferred.py b/weasyprint/layout/preferred.py
index 917e42705..c508399de 100644
--- a/weasyprint/layout/preferred.py
+++ b/weasyprint/layout/preferred.py
@@ -289,10 +289,10 @@ def inline_line_widths(context, box, outer, is_line_start, minimum,
             else:
                 (skip, skip_stack), = skip_stack.items()
                 assert skip_stack is None
-            child_text = child.text[(skip or 0):]
+            child_text = child.text.encode()[(skip or 0):]
             if is_line_start and space_collapse:
-                child_text = child_text.lstrip(' ')
-            if minimum and child_text == ' ':
+                child_text = child_text.lstrip(b' ')
+            if minimum and child_text == b' ':
                 lines = [0, 0]
             else:
                 max_width = 0 if minimum else None
@@ -302,8 +302,8 @@ def inline_line_widths(context, box, outer, is_line_start, minimum,
                     resume_index += new_resume_index
                     _, _, new_resume_index, width, _, _ = (
                         split_first_line(
-                            child_text[resume_index:], child.style, context,
-                            max_width, child.justification_spacing,
+                            child_text[resume_index:].decode(), child.style,
+                            context, max_width, child.justification_spacing,
                             is_line_start=is_line_start, minimum=True))
                     lines.append(width)
                     if first_line: