Work diff canvas splits hunk lines at form feeds and Unicode line separators #281

Expected and actual behavior / 实际现象与期望

parse_unified_diff (server/unified_diff.py) should treat each patch line as Git emitted it. It iterates over source.splitlines(), and str.splitlines() breaks not only at \n but also at form feed (\f), vertical tab, \x1c–\x1e, \x85, U+2028 and U+2029. Such characters inside a source line split that line in two:

• the rest of the line becomes a separate patch line; if it starts with -, + or a space, it is rendered as a phantom removed/added/context line, and the additions/deletions counts and every following line number are wrong;
• otherwise the rest is silently dropped, so the shown line is truncated.

Form feeds are common page separators in Python, C and Lisp sources, and U+2028 appears in JavaScript/JSON strings.

Steps to reproduce / 复现步骤

In a Git repository, commit app.py containing:

def a():
    pass
<form feed>
def b():
    pass

Change it to return 1 in a() and replace the last line with x = "a<form feed>-b", then run:

import subprocess
from server.unified_diff import parse_unified_diff

patch = subprocess.run(["git", "diff", "HEAD", "--"], capture_output=True, text=True).stdout
file = parse_unified_diff(patch)["files"][0]
print(file["additions"], file["deletions"], file["hunks"][0]["lines"][-2:])

Actual: 2 3. The added line is shown as x = "a, followed by a phantom removed line b" at old line 6.
Expected: 2 2, with the added line x = "a\f-b".

Relevant environment / 使用环境

Platform-independent; current main (fdd672e).

Sanitized evidence / 错误信息或截图（请先脱敏）

Git terminates patch lines with \n only. Splitting on \n and removing a trailing \r keeps the current handling of CRLF content and CRLF-framed patches.

Related issues or fixes / 相关问题或修复

#274 / #275 fixed a different parser defect in the same function.