fix: fix lots of position off-by-ones

Previously the PosTrackingReader always mysteriously subtracted 1
from the current position ... this wasn't sound at all ... the machine
just happens to often call `Tokenizer::unread_char` ... but not always.
E.g. for proper comments it didn't which resulted in their offset and
spans being off-by-one, which is fixed by this commit (see test_spans.rs).

1 files changed, 4 insertions, 4 deletions

diff --git a/src/tokenizer.rs b/src/tokenizer.rs
index 02a4d62..e8a8908 100644
--- a/src/tokenizer.rs
+++ b/src/tokenizer.rs
@@ -106,7 +106,7 @@ impl From<State> for InternalState {
     }
 }
 
-impl<R: Reader + Position<O>, O, E: Emitter<O>> Tokenizer<R, O, E> {
+impl<R: Reader + Position<O>, O: Offset, E: Emitter<O>> Tokenizer<R, O, E> {
     /// Test-internal function to override internal state.
     ///
     /// Only available with the `integration-tests` feature which is not public API.
@@ -123,7 +123,7 @@ impl<R: Reader + Position<O>, O, E: Emitter<O>> Tokenizer<R, O, E> {
     /// Just a helper method for the machine.
     #[inline]
     pub(crate) fn emit_error(&mut self, error: Error) {
-        self.emitter.emit_error(error, self.reader.position());
+        self.emitter.emit_error(error, self.reader.position() - 1);
     }
 
     /// Assuming the _current token_ is an end tag, return true if all of these hold. Return false otherwise.
@@ -140,14 +140,14 @@ impl<R: Reader + Position<O>, O, E: Emitter<O>> Tokenizer<R, O, E> {
 
     #[inline]
     pub(crate) fn init_start_tag(&mut self) {
-        self.emitter.init_start_tag(self.reader.position());
+        self.emitter.init_start_tag(self.reader.position() - 1);
         self.current_tag_name.clear();
         self.is_start_tag = true;
     }
 
     #[inline]
     pub(crate) fn init_end_tag(&mut self) {
-        self.emitter.init_end_tag(self.reader.position());
+        self.emitter.init_end_tag(self.reader.position() - 1);
         self.current_tag_name.clear();
         self.is_start_tag = false;
     }