From 3414919e32b4011a9673c568d98250967595cfa6 Mon Sep 17 00:00:00 2001
From: Jan Schlosser <jan@steffiundjan.de>
Date: Tue, 20 Oct 2026 09:41:11 +0200
Subject: [PATCH] only ignore whitespace between adjacent encoded words
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The whitespace stripping added in #168 was unanchored and removed
whitespace before every `=?` in the remainder, so plain text followed
by an encoded word lost its separating space (e.g. "a x =?..?=" became
"a xb"). Per RFC 2047 §6.2, only linear white space between two
adjacent encoded words is ignored.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 lib/mail/parsers/rfc_2822.ex        |  3 +-
 test/mail/parsers/rfc_2822_test.exs | 49 +++++++++++++++++++++++++++++
 2 files changed, 51 insertions(+), 1 deletion(-)

diff --git a/lib/mail/parsers/rfc_2822.ex b/lib/mail/parsers/rfc_2822.ex
index e4d7df72..b470928b 100644
--- a/lib/mail/parsers/rfc_2822.ex
+++ b/lib/mail/parsers/rfc_2822.ex
@@ -540,7 +540,8 @@ defmodule Mail.Parsers.RFC2822 do
         decoded_string = charset_handler.(charset, decoded_string)
 
         # Remove space if immediately followed by another encoded word string
-        remainder = Regex.replace(~r/\s+\=\?/, remainder, "=?")
+        remainder =
+          Regex.replace(~r/\A(?:(?:\r\n)?[ \t])+(?==\?[^?]+\?[BbQq]\?[^?]*\?=)/, remainder, "")
 
         decoded_string <> parse_encoded_word(remainder, opts)
 
diff --git a/test/mail/parsers/rfc_2822_test.exs b/test/mail/parsers/rfc_2822_test.exs
index 515cad30..e3040206 100644
--- a/test/mail/parsers/rfc_2822_test.exs
+++ b/test/mail/parsers/rfc_2822_test.exs
@@ -546,6 +546,55 @@ defmodule Mail.Parsers.RFC2822Test do
     assert message.headers["x-encoded-word"] == "hεłłø wørłd"
   end
 
+  test "ignores whitespace between adjacent encoded words (RFC 2047 §6.2)" do
+    message =
+      parse_email("""
+      X-Adjacent: =?utf-8?Q?a?= =?utf-8?Q?b?=
+      X-Folded-Space: =?utf-8?Q?a?=
+       =?utf-8?Q?b?=
+      X-Folded-Tab: =?utf-8?Q?a?=
+      \t=?utf-8?Q?b?=
+
+      Body
+      """)
+
+    assert message.headers["x-adjacent"] == "ab"
+    assert message.headers["x-folded-space"] == "ab"
+    assert message.headers["x-folded-tab"] == "ab"
+  end
+
+  test "only ignores linear white space between adjacent encoded words (RFC 2047 §6.2)" do
+    message =
+      Mail.Parsers.RFC2822.parse(
+        "X-LF: =?utf-8?Q?a?=\n=?utf-8?Q?b?=\r\n" <>
+          "X-CR: =?utf-8?Q?a?=\r=?utf-8?Q?b?=\r\n" <>
+          "X-VT: =?utf-8?Q?a?=\v=?utf-8?Q?b?=\r\n" <>
+          "X-FF: =?utf-8?Q?a?=\f=?utf-8?Q?b?=\r\n" <>
+          "\r\nBody\r\n"
+      )
+
+    assert message.headers["x-lf"] == "a\nb"
+    assert message.headers["x-cr"] == "a\rb"
+    assert message.headers["x-vt"] == "a\vb"
+    assert message.headers["x-ff"] == "a\fb"
+  end
+
+  test "keeps whitespace between encoded word and plain text (RFC 2047 §6.2)" do
+    message =
+      parse_email("""
+      X-Mixed: =?utf-8?Q?a?= x =?utf-8?Q?b?=
+      X-Stray: =?utf-8?Q?a?= =? x
+      Subject: =?utf-8?Q?Gr=C3=BC=C3=9Fe?= from the
+       =?utf-8?Q?M=C3=BCller?= family
+
+      Body
+      """)
+
+    assert message.headers["x-mixed"] == "a x b"
+    assert message.headers["x-stray"] == "a =? x"
+    assert message.headers["subject"] == "Grüße from the Müller family"
+  end
+
   test "allow empty body (RFC2822 §3.5)" do
     message =
       parse_email("""