[clang-format] output-replacements-xml returns non empty content for previously formatted code with function-try-block

[Alexolut]

Consider following C++ code with function-try-block:

void TestFunc(const std::string &data) try {
  do_something();
  do_something_more();
} catch (const std::exception &ex) {
  log_error(ex);
}

Given .clang-format:

---
Language:        Cpp
KeepEmptyLinesAtTheStartOfBlocks: false
SeparateDefinitionBlocks: Always
...

I found that after applying code formatting to this source file, e.g.

clang-format --style=file -i file.cpp

The following call

clang-format --style=file --output-replacements-xml file.cpp

returns non-empty diff:

<?xml version='1.0'?>
<replacements xml:space='preserve' incomplete_format='false'>
<replacement offset='125' length='4'>&#13;&#10;  </replacement>
</replacements>

As far as I understand there should be no replacement offset section in result if I did formatting before with the same settings.

Investigation of that issue started here. I use clang-format 18.1.7 on Windows.

[2bndy5]

Given the file.cpp and clang-format config, I was able to reproduce this using

• v18.1.8 on Ubuntu (actually using WSL on Windows).
• v18.1.7 on Windows

Note

Contrary to steps described in OP, I did not need to first format the file.cpp before generating the XML replacements.

<?xml version='1.0'?>
<replacements xml:space='preserve' incomplete_format='false'>
<replacement offset='125' length='4'>&#13;&#10;  </replacement>
</replacements>

which (when the replacement is applied) translates to a useless diff:

 void TestFunc(const std::string &data) try {
   do_something();
   do_something_more();
-} catch (const std::exception &ex) {
-  log_error(ex);
+} catch (const std::exception &ex) {
+  log_error(ex);
 }

Proof using python

from pathlib import Path
import xml.etree.ElementTree as ET
xml = "<?xml version='1.0'?><replacements xml:space='preserve' incomplete_format='false'><replacement offset='125' length='4'>&#13;&#10;  </replacement></replacements>"
tree = ET.fromstring(xml)
offset = int(tree[0].attrib["offset"])
length = int(tree[0].attrib["length"])
replacement = tree[0].text
f = Path("file.cpp")
src = f.read_bytes()[offset : offset + length].decode(encoding="utf-8")
assert src == replacement

This doesn't seem specific to line endings as I was able to reproduce this behavior with either LF or CRLF (on both Windows and Ubuntu).