From 31d7bbf622fc701552cc72678ccf2c56165eaf9a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jonas=20Sch=C3=A4fer?= <jonas@zombofant.net>
Date: Fri, 18 Apr 2025 19:42:42 +0200
Subject: [PATCH] xso: discard xml:lang implicitly

Because this attribute may occur in random places, it makes no sense
failing on it. We discard it after attribute processing though, so it
can still be captured by structs which are explicitly interested in it.
---
 parsers/src/util/macro_tests.rs | 17 +++++++++++++++++
 xso-proc/src/compound.rs        |  6 +++++-
 xso/ChangeLog                   |  5 +++++
 3 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/parsers/src/util/macro_tests.rs b/parsers/src/util/macro_tests.rs
index 70bbffb9..a477ed94 100644
--- a/parsers/src/util/macro_tests.rs
+++ b/parsers/src/util/macro_tests.rs
@@ -139,6 +139,23 @@ fn empty_unexpected_attribute() {
     }
 }
 
+#[test]
+#[cfg_attr(
+    feature = "disable-validation",
+    should_panic = "unexpected result: Ok("
+)]
+fn empty_ignores_xml_lang() {
+    #[allow(unused_imports)]
+    use core::{
+        option::Option::{None, Some},
+        result::Result::{Err, Ok},
+    };
+    match parse_str::<Empty>("<foo xmlns='urn:example:ns1' xml:lang='bar'/>") {
+        Ok(Empty) => (),
+        other => panic!("unexpected result: {:?}", other),
+    }
+}
+
 #[test]
 #[cfg_attr(
     feature = "disable-validation",
diff --git a/xso-proc/src/compound.rs b/xso-proc/src/compound.rs
index bf149119..d91fa8c3 100644
--- a/xso-proc/src/compound.rs
+++ b/xso-proc/src/compound.rs
@@ -396,7 +396,11 @@ impl Compound {
             }
         }
 
-        let mut discard_attr = TokenStream::default();
+        // We always implicitly discard the `xml:lang` attribute. Its
+        // processing is handled using the `#[xml(language)]` meta.
+        let mut discard_attr = quote! {
+            let _ = #attrs.remove(::xso::exports::rxml::Namespace::xml(), "lang");
+        };
         for (xml_namespace, xml_name) in self.discard_attr.iter() {
             let xml_namespace = match xml_namespace {
                 Some(v) => v.to_token_stream(),
diff --git a/xso/ChangeLog b/xso/ChangeLog
index f00527a6..002c8236 100644
--- a/xso/ChangeLog
+++ b/xso/ChangeLog
@@ -12,6 +12,11 @@ Version NEXT:
 
         This change overall allows for more flexibility in the implementation
         of text codecs.
+      - `xml:lang` attributes are now silently discarded during
+        deserialisation unless captured by `#[xml(attribute = "xml:lang")]` or
+        equivalent. This is because `xml:lang` is special as per
+        XML 1.0 § 2.12 and it is generally preferable to allow it to occur
+        anywhere in the document.
     * Added
       - Support for child elements in derive macros. Child elements may also
         be wrapped in Option or Box or in containers like Vec or HashSet.