Updated changelog.
diff --git a/changelog.txt b/changelog.txt
index b4ee8ea..9605c28 100644
--- a/changelog.txt
+++ b/changelog.txt
@@ -1,3 +1,147 @@
+[0.21.0]
+
+ * Updated to version 0.21 of spec.
+ * Added latex renderer (#31). New exported function in API:
+ `cmark_render_latex`. New source file: `src/latex.hs`.
+ * Updates for new HTML block spec. Removed old `html_block_tag` scanner.
+ Added new `html_block_start` and `html_block_start_7`, as well
+ as `html_block_end_n` for n = 1-5. Rewrote block parser for new HTML
+ block spec.
+ * We no longer preprocess tabs to spaces before parsing.
+ Instead, we keep track of both the byte offset and
+ the (virtual) column as we parse block starts.
+ This allows us to handle tabs without converting
+ to spaces first. Tabs are left as tabs in the output, as
+ per the revised spec.
+ * Removed utf8 validation by default. We now replace null characters
+ in the line splitting code.
+ * Added `CMARK_OPT_VALIDATE_UTF8` option and command-line option
+ `--validate-utf8`. This option causes cmark to check for valid
+ UTF-8, replacing invalid sequences with the replacement
+ character, U+FFFD. Previously this was done by default in
+ connection with tab expansion, but we no longer do it by
+ default with the new tab treatment. (Many applications will
+ know that the input is valid UTF-8, so validation will not
+ be necessary.)
+ * Added `CMARK_OPT_SAFE` option and `--safe` command-line flag.
+ + Added `CMARK_OPT_SAFE`. This option disables rendering of raw HTML
+ and potentially dangerous links.
+ + Added `--safe` option in command-line program.
+ + Updated `cmark.3` man page.
+ + Added `scan_dangerous_url` to scanners.
+ + In HTML, suppress rendering of raw HTML and potentially dangerous
+ links if `CMARK_OPT_SAFE`. Dangerous URLs are those that begin
+ with `javascript:`, `vbscript:`, `file:`, or `data:` (except for
+ `image/png`, `image/gif`, `image/jpeg`, or `image/webp` mime types).
+ + Added `api_test` for `OPT_CMARK_SAFE`.
+ + Rewrote `README.md` on security.
+ * Limit ordered list start to 9 digits, per spec.
+ * Added width parameter to `render_man` (API change).
+ * Extracted common renderer code from latex, man, and commonmark
+ renderers into a separate module, `renderer.[ch]` (#63). To write a
+ renderer now, you only need to write a character escaping function
+ and a node rendering function. You pass these to `cmark_render`
+ and it handles all the plumbing (including line wrapping) for you.
+ So far this is an internal module, but we might consider adding
+ it to the API in the future.
+ * commonmark writer: correctly handle email autolinks.
+ * commonmark writer: escape `!`.
+ * Fixed soft breaks in commonmark renderer.
+ * Fixed scanner for link url. re2c returns the longest match, so we
+ were getting bad results with `[link](foo\(and\(bar\)\))`
+ which it would parse as containing a bare `\` followed by
+ an in-parens chunk ending with the final paren.
+ * Allow non-initial hyphens in html tag names. This allows for
+ custom tags, see jgm/CommonMark#239.
+ * Updated `test/smart_punct.txt`.
+ * Implemented new treatment of hyphens with `--smart`, converting
+ sequences of hyphens to sequences of em and en dashes that contain no
+ hyphens.
+ * HTML renderer: properly split info on first space char (see
+ jgm/commonmark.js#54).
+ * Changed version variables to functions (#60, Andrius Bentkus).
+ This is easier to access using ffi, since some languages, like C#
+ like to use only function interfaces for accessing library
+ functionality.
+ * `process_emphasis`: Fixed setting lower bound to potential openers.
+ Renamed `potential_openers` -> `openers_bottom`.
+ Renamed `start_delim` -> `stack_bottom`.
+ * Added case for #59 to `pathological_test.py`.
+ * Fixed emphasis/link parsing bug (#59).
+ * Fixed off-by-one error in line splitting routine.
+ This caused certain NULLs not to be replaced.
+ * Don't rtrim in `subject_from_buffer`. This gives bad results in
+ parsing reference links, where we might have trailing blanks
+ (`finalize` removes the bytes parsed as a reference definition;
+ before this change, some blank bytes might remain on the line).
+ + Added `column` and `first_nonspace_column` fields to `parser`.
+ + Added utility function to advance the offset, computing
+ the virtual column too. Note that we don't need to deal with
+ UTF-8 here at all. Only ASCII occurs in block starts.
+ + Significant performance improvement due to the fact that
+ we're not doing UTF-8 validation.
+ * Fixed entity lookup table. The old one had many errors.
+ The new one is derived from the list in the npm entities package.
+ Since the sequences can now be longer (multi-code-point), we
+ have bumped the length limit from 4 to 8, which also affects
+ `houdini_html_u.c`. An example of the kind of error that was fixed:
+ `≧̸` should be rendered as "≧̸" (U+02267 U+00338), but it was
+ being rendered as "≧" (which is the same as `≧`).
+ * Replace gperf-based entity lookup with binary tree lookup.
+ The primary advantage is a big reduction in the size of
+ the compiled library and executable (> 100K).
+ There should be no measurable performance difference in
+ normal documents. I detected only a slight performance
+ hit in a file containing 1,000,000 entities.
+ + Removed `src/html_unescape.gperf` and `src/html_unescape.h`.
+ + Added `src/entities.h` (generated by `tools/make_entities_h.py`).
+ + Added binary tree lookup functions to `houdini_html_u.c`, and
+ use the data in `src/entities.h`.
+ * Renamed `entities.h` -> `entities.inc`, and
+ `tools/make_entities_h.py` -> `tools/make_entitis_inc.py`.
+ * Fixed cases like
+ ```
+ [ref]: url
+ "title" ok
+ ```
+ Here we should parse the first line as a reference.
+ * `inlines.c`: Added utility functions to skip spaces and line endings.
+ * Fixed backslashes in link destinations that are not part of escapes
+ (jgm/commonmark#45).
+ * `process_line`: Removed "add newline if line doesn't have one."
+ This isn't actually needed.
+ * Small logic fixes and a simplification in `process_emphasis`.
+ * Added more pathological tests:
+ + Many link closers with no openers.
+ + Many link openers with no closers.
+ + Many emph openers with no closers.
+ + Many closers with no openers.
+ + `"*a_ " * 20000`.
+ * Fixed `process_emphasis` to handle new pathological cases.
+ Now we have an array of pointers (`potential_openers`),
+ keyed to the delim char. When we've failed to match a potential opener
+ prior to point X in the delimiter stack, we reset `potential_openers`
+ for that opener type to X, and thus avoid having to look again through
+ all the openers we've already rejected.
+ * `process_inlines`: remove closers from delim stack when possible.
+ When they have no matching openers and cannot be openers themselves,
+ we can safely remove them. This helps with a performance case:
+ `"a_ " * 20000` (jgm/commonmark.js#43).
+ * Roll utf8proc_charlen into utf8proc_valid (Nick Wellnhofer).
+ Speeds up "make bench" by another percent.
+ * `spec_tests.py`: allow `→` for tab in HTML examples.
+ * `normalize.py`: don't collapse whitespace in pre contexts.
+ * Use utf-8 aware re2c.
+ * Makefile afl target: removed `-m none`, added `CMARK_OPTS`.
+ * README: added `make afl` instructions.
+ * Limit generated generated `cmark.3` to 72 character line width.
+ * Travis: switched to containerized build system.
+ * Removed `debug.h`. (It uses GNU extensions, and we don't need it anyway.)
+ * Removed sundown from benchmarks, because the reading was anomalous.
+ sundown had an arbitrary 16MB limit on buffers, and the benchmark
+ input exceeded that. So who knows what we were actually testing?
+ Added hoedown, sundown's successor, which is a better comparison.
+
[0.20.0]
* Fixed bug in list item parsing when items indented >= 4 spaces (#52).
