Previously, we would insert characters into the temporary buffer for two
reasons:
To keep track of text we'd seen and moved beyond but had not emitted
as character tokens in order to emit them later; and
Record strings for comments and doctypes.
Use 1 was a bit silly because in order to get the correct token
positions, we would mark the input stream when clearing the temporary
buffer, reset the input to the mark, and then advance the input and a
pointer into the temporary buffer in lock step, emitting character
tokens.
Now we just mark the input stream, and then begin emitting from the mark
point as needed.
This has the advantage that it frees us to use the temporary buffer for
recording the escaped script tag rather than the _script_data_buffer
which is now removed.
Previously, we would insert characters into the temporary buffer for two reasons:
Use 1 was a bit silly because in order to get the correct token positions, we would mark the input stream when clearing the temporary buffer, reset the input to the mark, and then advance the input and a pointer into the temporary buffer in lock step, emitting character tokens.
Now we just mark the input stream, and then begin emitting from the mark point as needed.
This has the advantage that it frees us to use the temporary buffer for recording the escaped
scripttag rather than the_script_data_bufferwhich is now removed.