Don't insert hard breaks inside multi-line inline HTML tags

CHANGES.md

@@ -26,6 +26,7 @@
 - [pull #737] Fix XSS from code spans that the span gamut doesn't treat as code spans in safe mode (#726)
 - [pull #743] Add a `strip-syntax` option to the `header-ids` extra that leaves link URLs, image paths, inline HTML tags and emphasis markers out of header ids (#192)
 - [pull #747] Fix `link-patterns` extra linking text inside Markdown links and images whose label wraps onto another line
+- [pull #755] Fix hard breaks and the `breaks` extra inserting `<br />` inside inline HTML tags and link titles that span several lines
 
 
 ## python-markdown2 2.5.5

lib/markdown2.py

@@ -1357,7 +1357,32 @@ def _run_span_gamut(self, text: str) -> str:
         text = self._do_italics_and_bold(text)
 
         # Do hard breaks
-        text = re.sub(r" {2,}\n(?!\<(?:\/?(ul|ol|li))\>)", "<br%s\n" % self.empty_element_suffix, text)
+        text = self._do_hard_breaks(text)
+
+        return text
+
+    def _do_hard_breaks(self, text: str, pattern: str = r' {2,}') -> str:
+        '''
+        Replace line endings matching `pattern` with `<br>` tags. Line endings
+        inside a multi-line HTML tag (eg: an `<img>` with its attributes on
+        separate lines) are left alone, as a `<br>` there would break the tag.
+        '''
+        tags = {}
+
+        def hash_tag(match: re.Match[str]) -> str:
+            if '\n' not in match.group(0):
+                return match.group(0)
+            key = _hash_text(match.group(0))
+            tags[key] = match.group(0)
+            return key
+
+        if '<' in text and '\n' in text:  # guard for perf
+            text = self._sorta_html_tokenize_re.sub(hash_tag, text)
+
+        text = re.sub(pattern + r"\n(?!\<(?:\/?(ul|ol|li))\>)", "<br%s\n" % self.empty_element_suffix, text)
+
+        for key, tag in tags.items():
+            text = text.replace(key, tag)
 
         return text
 
@@ -3575,10 +3600,7 @@ def run(self, text: str):
         else:
             pattern = r' {2,}'
 
-        break_tag = "<br%s\n" % self.md.empty_element_suffix
-        text = re.sub(pattern + r"\n(?!\<(?:\/?(ul|ol|li))\>)", break_tag, text)
-
-        return text
+        return self.md._do_hard_breaks(text, pattern)
 
 
 class CodeFriendly(GFMItalicAndBoldProcessor):

test/tm-cases/break_on_newline_html_tags.html

@@ -0,0 +1,7 @@
+<p>An image <img src="/path/to/cat.png"
+    alt="A cat"> sits on this line,<br />
+and this line still gets a break.</p>
+
+<p>A raw <a href="https://example.com/"
+title="Example">link</a> and a <a href="https://example.com/" title="A title
+that wraps">Markdown link</a> keep their attributes.</p>

test/tm-cases/break_on_newline_html_tags.opts

@@ -0,0 +1 @@
+{"extras": ["break-on-newline"]}

test/tm-cases/break_on_newline_html_tags.text

@@ -0,0 +1,7 @@
+An image <img src="/path/to/cat.png"
+    alt="A cat"> sits on this line,
+and this line still gets a break.
+
+A raw <a href="https://example.com/"
+title="Example">link</a> and a [Markdown link](https://example.com/ "A title
+that wraps") keep their attributes.

test/tm-cases/hard_break_html_tags.html

@@ -0,0 +1,3 @@
+<p>An image <img src="/path/to/cat.png"  
+    alt="A cat"> with trailing spaces inside the tag,<br />
+and a hard break after it.</p>

test/tm-cases/hard_break_html_tags.text

@@ -0,0 +1,3 @@
+An image <img src="/path/to/cat.png"  
+    alt="A cat"> with trailing spaces inside the tag,  
+and a hard break after it.