From 02998950d60e03d9e61ac573422b724e4578df06 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=C3=93tz=C3=AF=20f=C3=B3n=20Gl=C3=BC=CF=B7gor?=
 <141450226+Otzie2023@users.noreply.github.com>
Date: Thu, 3 Sep 2026 20:47:38 +0200
Subject: [PATCH 1/2] Describe how non-ASCII variable names behave

The note claimed PHP "doesn't support Unicode variable names" and then
explained that they work anyway. Replace it with a description of the
observable behaviour.
---
 language/variables.xml | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/language/variables.xml b/language/variables.xml
index 710a69bf4224..21f40343ba2f 100644
--- a/language/variables.xml
+++ b/language/variables.xml
@@ -23,10 +23,21 @@
 
    <note>
     <simpara>
-     PHP doesn't support Unicode variable names, however, some character
-     encodings (such as UTF-8) encode characters in such a way that all bytes
-     of a multi-byte character fall within the allowed range, thus making it a
-     valid variable name.
+     Variable names are compared as bytes, not as characters. PHP does not
+     interpret or validate the encoding of the bytes from 128 through 255, and
+     two names refer to the same variable only when their bytes are identical.
+    </simpara>
+    <simpara>
+     One consequence is that names written in UTF-8 work, because every byte of
+     a UTF-8 multi-byte sequence falls inside the accepted range. So do names
+     written in other encodings, and so do byte sequences that are not valid
+     text in any encoding.
+    </simpara>
+    <simpara>
+     Another is that two names which display identically may be distinct.
+     <literal>ä</literal> encoded as U+00E4 and as <literal>a</literal>
+     followed by the combining diaeresis U+0308 are different variables, as are
+     a name with and without a trailing U+00A0 NO-BREAK SPACE.
     </simpara>
    </note>
 

From 5930719979cf5be786cda4f45f28802e764aaf34 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=C3=93tz=C3=AF=20f=C3=B3n=20Gl=C3=BC=CF=B7gor?=
 <141450226+Otzie2023@users.noreply.github.com>
Date: Fri, 4 Sep 2026 09:45:14 +0200
Subject: [PATCH 2/2] Update language/variables.xml

Co-authored-by: Louis-Arnaud <la.catoire@gmail.com>
---
 language/variables.xml | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/language/variables.xml b/language/variables.xml
index 21f40343ba2f..95d21540ef85 100644
--- a/language/variables.xml
+++ b/language/variables.xml
@@ -29,9 +29,10 @@
     </simpara>
     <simpara>
      One consequence is that names written in UTF-8 work, because every byte of
-     a UTF-8 multi-byte sequence falls inside the accepted range. So do names
-     written in other encodings, and so do byte sequences that are not valid
-     text in any encoding.
+     a UTF-8 multi-byte sequence falls inside the accepted range. The same holds
+     for any encoding whose non-ASCII bytes all fall in that range, and for byte
+     sequences that are not valid text in any encoding. Encodings that use bytes
+     below 128 within a multi-byte sequence, such as Shift-JIS or UTF-16, do not.
     </simpara>
     <simpara>
      Another is that two names which display identically may be distinct.