From e6f5736fdcc0288463f8cdb92a566e15a2372433 Mon Sep 17 00:00:00 2001
From: hanjinpeng <hanjinpeng@kylinos.cn>
Date: Thu, 17 Sep 2026 02:31:13 -0400
Subject: [PATCH] utf7.c: decode "+-" as a plus sign

RFC 2152 specifies that "+-" encodes a plus sign.  The decoder saw no
Base64 character after the `+', then swallowed the `-' as the optional
end of a shifted sequence, so every literal plus sign was silently lost:

  $ echo '1 +- 1 = 2' | recode utf-7..utf-8
  1  1 = 2

Output the plus sign.  Also encode a plus sign as "+-", as recommended
by RFC 2152, rather than as the longer, still valid, "+ACs-".

Add a test.
---
 src/utf7.c        | 16 ++++++++++++++++
 tests/t40_utf7.py |  5 ++++-
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/src/utf7.c b/src/utf7.c
index 745d288..65bec3b 100644
--- a/src/utf7.c
+++ b/src/utf7.c
@@ -85,6 +85,15 @@ transform_utf16_utf7 (RECODE_SUBTASK subtask)
 	if (!recode_get_ucs2 (&value, subtask))
 	  SUBTASK_RETURN (subtask);
       }
+    else if (value == '+')
+      {
+	/* A plus sign is simply written as "+-".  */
+
+	recode_put_byte ('+', subtask);
+	recode_put_byte ('-', subtask);
+	if (!recode_get_ucs2 (&value, subtask))
+	  SUBTASK_RETURN (subtask);
+      }
     else
       {
 	/* Copy a string of non-direct characters.  */
@@ -162,6 +171,13 @@ transform_utf7_utf16 (RECODE_SUBTASK subtask)
     if (character == '+')
       {
 	character = recode_get_byte (subtask);
+	if (character == '-')
+	  {
+	    /* "+-" is the encoding of a plus sign itself.  */
+	    recode_put_ucs2 ('+', subtask);
+	    character = recode_get_byte (subtask);
+	    continue;
+	  }
 	while (IS_BASE64 (character))
 	  {
 	    /* Process first byte of first quadruplet.  */
diff --git a/tests/t40_utf7.py b/tests/t40_utf7.py
index 329b617..ee091d0 100644
--- a/tests/t40_utf7.py
+++ b/tests/t40_utf7.py
@@ -8,6 +8,7 @@
     'Hi Mom +Jjo!\n',
     '+ZeVnLIqe\n',
     'Item 3 is +AKM-1.\n',
+    '1 +- 1 = 2\n',
     ]
 
 outputs = [
@@ -17,7 +18,9 @@
      '0x263A, 0x0021, 0x000A\n'),
     '0xFEFF, 0x65E5, 0x672C, 0x8A9E, 0x000A\n',
     ('0xFEFF, 0x0049, 0x0074, 0x0065, 0x006D, 0x0020, 0x0033, 0x0020,\n'
-     '0x0069, 0x0073, 0x0020, 0x00A3, 0x0031, 0x002E, 0x000A\n')]
+     '0x0069, 0x0073, 0x0020, 0x00A3, 0x0031, 0x002E, 0x000A\n'),
+    ('0xFEFF, 0x0031, 0x0020, 0x002B, 0x0020, 0x0031, 0x0020, 0x003D,\n'
+     '0x0020, 0x0032, 0x000A\n')]
 
 class Test: