Forbid whitespace before XML prolog

kornelski · kornelski · commit 535914eaf8ec · 2023-05-22T17:02:59.000+01:00
diff --git a/src/reader/parser.rs b/src/reader/parser.rs
@@ -90,10 +90,11 @@ pub(crate) struct PullParser {
 #[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord)]
 enum Encountered {
     None = 0,
-    Declaration = 1,
-    Comment = 2,
-    Doctype = 3,
-    Element = 4,
+    AnyChars, // whitespace before <?xml is not allowed
+    Declaration,
+    Comment,
+    Doctype,
+    Element,
 }
 
 impl PullParser {
@@ -117,7 +118,7 @@ impl PullParser {
         PullParser {
             config,
             lexer,
-            st: State::OutsideTag,
+            st: State::DocumentStart,
             state_after_reference: State::OutsideTag,
             buf: String::new(),
             entities: HashMap::new(),
@@ -159,7 +160,7 @@ impl PullParser {
 
         // If declaration was not parsed and we have encountered an element,
         // emit this declaration as the next event.
-        if prev_enc < Encountered::Declaration {
+        if prev_enc == Encountered::None {
             self.push_pos();
             Some(Ok(XmlEvent::StartDocument {
                 version: DEFAULT_VERSION,
@@ -191,6 +192,7 @@ pub enum State {
     InsideDeclaration(DeclarationSubstate),
     InsideDoctype(DoctypeSubstate),
     InsideReference,
+    DocumentStart,
 }
 
 #[derive(Copy, Clone, PartialEq)]
@@ -423,6 +425,7 @@ impl PullParser {
             State::InsideProcessingInstruction(s) => self.inside_processing_instruction(t, s),
             State::InsideDoctype(s)               => self.inside_doctype(t, s),
             State::InsideDeclaration(s)           => self.inside_declaration(t, s),
+            State::DocumentStart                  => self.document_start(t),
         }
     }
 
diff --git a/src/reader/parser/outside_tag.rs b/src/reader/parser/outside_tag.rs
@@ -128,7 +128,6 @@ impl PullParser {
                     Token::ProcessingInstructionStart =>
                         self.into_state(State::InsideProcessingInstruction(ProcessingInstructionSubstate::PIInsideName), next_event),
 
-
                     Token::CDataStart if self.depth() > 0 => {
                         self.into_state(State::InsideCData, next_event)
                     },
@@ -138,4 +137,55 @@ impl PullParser {
             }
         }
     }
+
+    pub fn document_start(&mut self, t: Token) -> Option<Result> {
+        debug_assert!(self.encountered < Encountered::Declaration);
+
+        match t {
+            Token::Character(c) => {
+                let next_event = self.set_encountered(Encountered::AnyChars);
+
+                if !is_whitespace_char(c) {
+                    return Some(self.error(SyntaxError::UnexpectedTokenOutsideRoot(t)));
+                }
+                self.inside_whitespace = true;
+
+                // skip whitespace outside of the root element
+                if (self.config.c.trim_whitespace && self.buf.is_empty()) ||
+                    (self.depth() == 0 && self.config.c.ignore_root_level_whitespace) {
+                        return self.into_state(State::OutsideTag, next_event);
+                }
+
+                self.push_pos();
+                self.buf.push(c);
+                self.into_state(State::OutsideTag, next_event)
+            },
+
+            Token::CommentStart  => {
+                let next_event = self.set_encountered(Encountered::Comment);
+                self.into_state(State::InsideComment, next_event)
+            }
+
+            Token::OpeningTagStart => {
+                let next_event = self.set_encountered(Encountered::Element);
+                self.nst.push_empty();
+                self.into_state(State::InsideOpeningTag(OpeningTagSubstate::InsideName), next_event)
+            },
+
+            Token::DoctypeStart => {
+                let next_event = self.set_encountered(Encountered::Doctype);
+                // We don't have a doctype event so skip this position
+                // FIXME: update when we have a doctype event
+                self.next_pos();
+                self.into_state(State::InsideDoctype(DoctypeSubstate::Outside), next_event)
+            },
+
+            Token::ProcessingInstructionStart => {
+                self.push_pos();
+                self.into_state_continue(State::InsideProcessingInstruction(ProcessingInstructionSubstate::PIInsideName))
+            },
+
+            _ => Some(self.error(SyntaxError::UnexpectedToken(t))),
+        }
+    }
 }
diff --git a/tests/event_reader.rs b/tests/event_reader.rs
@@ -285,7 +285,7 @@ fn tabs_1() {
     test(
         b"\t<a>\t<b/></a>",
         br#"
-            |1:2 StartDocument(1.0, UTF-8)
+            |1:1 StartDocument(1.0, UTF-8)
             |1:2 StartElement(a)
             |1:6 StartElement(b)
             |1:6 EndElement(b)
diff --git a/tests/oasis.fail.txt b/tests/oasis.fail.txt
@@ -1,6 +1,5 @@
 o-p04pass1 p04pass1.xml   names with all valid ASCII characters, and one from each    other class in NameChar  ; 5:8 Element A.-:̀· prefix is unbound
 o-p05pass1 p05pass1.xml   various valid Name constructions  ; 2:8 Element A:._-0 prefix is unbound
-o-p01fail1 p01fail1.xml   S cannot occur before the prolog  
 o-p09fail1 p09fail1.xml   EntityValue excludes '%'  
 o-p09fail2 p09fail2.xml   EntityValue excludes '&'  
 o-p12fail1 p12fail1.xml   '"' excluded  
@@ -10,7 +9,6 @@ o-p12fail4 p12fail4.xml   '>' excluded
 o-p12fail5 p12fail5.xml   '<' excluded  
 o-p12fail6 p12fail6.xml   built-in entity refs excluded  
 o-p12fail7 p12fail7.xml   The public ID has a tab character, which is disallowed  
-o-p22fail1 p22fail1.xml   prolog must start with XML decl  
 o-p30fail1 p30fail1.xml   An XML declaration is not the same as a TextDecl  
 o-p31fail1 p31fail1.xml   external subset excludes doctypedecl  
 o-p32fail3 p32fail3.xml   initial S is required  
diff --git a/tests/sun-not-wf.fail.txt b/tests/sun-not-wf.fail.txt
@@ -27,7 +27,6 @@ pubid02 pubid02.xml   Illegal characters in public ID
 pubid03 pubid03.xml   Illegal characters in public ID
 pubid04 pubid04.xml   Illegal characters in public ID
 pubid05 pubid05.xml   SGML-ism: public ID without system ID
-sgml02 sgml02.xml   XML declaration must be at the very beginning of a document; it"s not a processing instruction
 sgml04 sgml04.xml   ATTLIST declarations apply to only one element, unlike SGML
 sgml05 sgml05.xml   ELEMENT declarations apply to only one element, unlike SGML
 sgml06 sgml06.xml   ATTLIST declarations are never global, unlike in SGML
diff --git a/tests/xmltest.fail.txt b/tests/xmltest.fail.txt
@@ -51,7 +51,6 @@ not-wf-sa-136 136.xml   Tag omission is invalid in XML.
 not-wf-sa-137 137.xml   Space is required before a content model. 
 not-wf-sa-138 138.xml   Invalid syntax for content particle. 
 not-wf-sa-139 139.xml   The element-content model should not be empty. 
-not-wf-sa-147 147.xml   XML Declaration may not be preceded by whitespace.
 not-wf-sa-149 149.xml   XML Declaration may not be within a DTD.
 not-wf-sa-158 158.xml   SGML-ism: "#NOTATION gif" can't have attributes. 
 not-wf-sa-159 159.xml   Uses '&' unquoted in an entity declaration,  which is illegal syntax for an entity reference.
