cmark/changelog.txt

[0.30.3]

  * Fix quadratic complexity bug with repeated `![[]()`.
    Resolves CVE-2023-22486. Add new pathological test. (John MacFarlane)

  * Allow declarations with no space, as per spec (#456, John MacFarlane).

  * Set `enumi*` counter correctly in LaTeX output (#451, John MacFarlane).

  * Allow `<!DOCTYPE` to be case-insensitive. (This conforms to the
    existing spec.) (John MacFarlane)

  * Fixed HTML comment scanning. Need to handle this case: `<!--> and -->`.
    Since the scanner finds the longest match, we had to
    move some of the logic outside of the scanner. (John MacFarlane)

  * Fix quadratic parsing issue with repeated `<!--` (this was not
    introduced by the previous fix, and not in a released version of cmark).
    Resolves CVE-2023-22484. Add new pathological test. (John MacFarlane)

  * Update HTML comment scanner to accord with commonmark/commonmark-spec#713
    (John MacFarlane).

  * Pathological tests: half the number of repetitions, and the timeout.
    This reduces the time needed for the pathological tests.
    (John MacFarlane)

  * Shrink `struct cmark_node` (#446). The `internal_offset` member is
    only used for headings and can be moved to `struct cmark_heading`.
    This reduces the size of `struct cmark_node` from 112 to 104 bytes on
    64-bit systems. (Nick Wellnhofer)

  * Add `-Wstrict-prototypes` and fix offending functions. (Nick
    Wellnhofer, Dan Cîrnaț)

  * Fix quadratic behavior involving `get_containing_block` (#431).
    Instead of searching for the containing block, update the tight list
    status when entering a child of a list item or exiting a list.
    (Nick Wellnhofer)

  * Fix `pathological_tests.py` (Nick Wellnhofer):
    - Use a multiprocessing.Queue to actually get results from spawned
      tests processes.
    - Fix the `allowed_failures` test.
    - Truncate actual output when printed.
    - Prepare for testing pathological behavior of the Commonmark renderer.

  * Fix source position bug with backticks (kyle).

[0.30.2]

  * Fix parsing of emphasis before links (#424, Nick Wellnhofer).
    Fixes a regression introduced with commit ed0a4bf.

  * Update to Unicode 14.0 (data-man).

  * Add `~` to safe href character set (#394, frogtile).

  * Update CMakeLists.txt (Saleem Abdulrasool).  Bump the minimum required
    CMake to 3.7. Imperatively define output name for static library.

  * Fix install paths in libcmark.pc (Sebastián Mancilla).
    `CMAKE_INSTALL_<dir>` can be relative or absolute path, so it is wrong to
    prefix CMAKE_INSTALL_PREFIX because if CMAKE_INSTALL_<dir> is set to an
    absolute path it will result in a malformed path with two absolute paths
    joined together.  Instead, use `CMAKE_INSTALL_FULL_<dir>` from
    GNUInstallDirs.

[0.30.1]

  * Properly indent block-level contents of list items in man (#258).
    This handles nested lists as well as items with multiple paragraphs.
    The change requires addition of a new field block_number_in_list_item
    to cmark_renderer, but this does not change the public API.
  * Fix quadratic behavior when parsing emphasis (#389, Nick
    Wellnhofer).  Delimiters can be deleted, so store delimiter positions
    instead of pointers in `openers_bottom`. Besides causing undefined
    behavior when reading a dangling pointer, this could also result
    in quadratic behavior when parsing emphasis.
  * Fix quadratic behavior when parsing smart quotes (#388, Nick Wellnhofer).
    Remove matching smart quote delimiters.  Otherwise, the same opener
    could be found over and over, preventing the `openers_bottom`
    optimization from kicking in and leading to quadratic behavior when
    processing lots of quotes.
  * Modify CMake configuration so that the project can be built with
    older versions of CMake (#384, Saleem Abdulrasool).  (In 0.30.0,
    some features were used that require CMake >= 3.3.) The cost of this
    backwards compatibility is that developers must now explicitly invoke
    `cmark_add_compile_options` when a new compilation target is added.
  * Remove a comma at the end of an enumerator list, which was flagged
    by clang as a C++11 extension.
  * make_man_page.py: use absolute path with CDLL. This avoids the error
    "file system relative paths not allowed in hardened programs."
  * Include cmark version in cmark(3) man page (instead of LOCAL).

[0.30.0]

  * Use official 0.30 spec.txt.
  * Add `cmark_get_default_mem_allocator()` (#330).  API change: this
    adds a new exported function in cmark.h.
  * Fix #383.  An optimization we used for emphasis parsing was
    too aggressive, causing us to miss some emphasis that was legal
    according to the spec.  We fix this by indexing the `openers_bottom`
    table not just by the type of delimiter and the length of the
    closing delimiter mod 3, but by whether the closing delimiter
    can also be an opener.  (The algorithm for determining emphasis
    matching depends on all these factors.)  Add regression test.
  * Fix quadratic behavior with inline HTML (#299, Nick Wellnhofer).
    Repeated starting sequences like `<?`, `<!DECL ` or `<![CDATA[` could
    lead to quadratic behavior if no matching ending sequence was found.
    Separate the inline HTML scanners. Remember if scanning the whole input
    for a specific ending sequence failed and skip subsequent scans.
  * Speed up hierarchy check in tree manipulation API (Nick Wellnhofer).
    Skip hierarchy check in the common case that the inserted child has
    no children.
  * Fix quadratic behavior when parsing inlines (#373, Nick Wellnhofer).
    The inline parsing code would call `cmark_node_append_child` to append
    nodes. This public function has a sanity check which is linear in the
    depth of the tree. Repeated calls could show quadratic behavior in
    degenerate trees. Use a special function to append nodes without this
    check.  (Issue found by OSS-Fuzz.)
  * Replace invalid characters in XML output (#365, Nick wellnhofer).
    Control characters, U+FFFE and U+FFFF aren't allowed in XML 1.0, so
    replace them with U+FFFD (replacement character). This doesn't solve
    the problem how to roundtrip these characters, but at least we don't
    produce invalid XML.
  * Avoid quadratic output growth with reference links (#354, Nick Wellnhofer).
    Keep track of the number bytes added through expansion of reference
    links and limit the total to the size of the input document. Always
    allow a minimum of 100KB.  Unfortunately, cmark has no error handling,
    so all we can do is to stop expanding reference links without returning
    an error. This should never be an issue in practice though. The 100KB
    minimum alone should cover all real-world cases.
  * Fix issue with type-7 HTML blocks interrupting paragraphs
    (see commonmark/commonmark.js#213).
  * Treat `textarea` like `script`, `style`, `pre` (type 1 HTML block),
    in accordance with spec change.
  * Define whitespace per spec (Asherah Conor).
  * Add `MAX_INDENT` for xml (#355).  Otherwise we can get quadratic
    increase in size with deeply nested structures.
  * Fix handling of empty strings when creating XML/HTML output
    (Steffen Kieß).
  * Commonmark renderer: always use fences for code (#317).
    This solves problems with adjacent code blocks being merged.
  * Improve rendering of commonmark code spans with spaces (#316).
  * Cleaner approach to max digits for numeric entities.
    This modifies unescaping in `houdini_html_u.c` rather than
    the entity handling in `inlines.c`.  Unlike the other,
    this approach works also in e.g. link titles.
  * Fix entity parser (and api test) to respect length limit on
    numeric entities.
  * Don't allow link destinations with unbalanced unescaped parentheses.
    See commonmark/commonmark.js#177.
  * `print_usage()`: Minor grammar fix, swap two words (#305, Øyvind A. Holm).
  * Don't call `memcpy` with `NULL` as first parameter.
    This is illegal according to the C standard, sec. 7.1.4.
    See <https://www.imperialviolet.org/2016/06/26/nonnull.html>.
  * Add needed include in `blocks.c`.
  * Fix unnecessary variable assignment.
  * Skip UTF-8 BOM if present at beginning of buffer (#334).
  * Fix URL check in `is_autolink` (Nick Wellnhofer).  In a recent commit,
    the check was changed to `strcmp`, but we really have to use `strncmp`.
  * Fix null pointer deref in `is_autolink` (Nick Wellnhofer).
    Introduced by a recent commit. Found by OSS-Fuzz.
  * Rearrange struct cmark_node (Nick Wellnhofer).  Introduce multi-purpose
    data/len members in struct cmark_node. This is mainly used to store
    literal text for inlines, code and HTML blocks.
    Move the content strbuf for blocks from `cmark_node` to `cmark_parser`.
    When finalizing nodes that allow inlines (paragraphs and headings),
    detach the strbuf and store the block content in the node's data/len
    members. Free the block content after processing inlines.
    Reduces size of struct `cmark_node` by 8 bytes.
  * Improve packing of `struct cmark_list` (Nick Wellnhofer).
  * Use C string instead of chunk in a number of contexts (Nick Wellnhofer,
    #309).  The node struct never references memory of other nodes now.
    Node accessors don't have to check for delayed creation of C strings,
    so parsing and iterating all literals using the public API should
    actually be faster than before.  These changes also reduce the size
    of `struct cmark_node`.
  * Add casts for MSVC10 (from kivikakk in cmark-cfm).
  * commonmark renderer:  better escaping in smart mode.  When
    `CMARK_OPT_SMART` is enabled, we escape literal `-`, `.`, and quote
    characters when needed to avoid their being "smartified."
  * Add options field to `cmark_renderer`.
  * commonmark.c - use `size_t` instead of `int`.
  * Include `string.h` in `cmark-fuzz.c`.
  * Fix #220 (hash collisions for references) (Vicent Marti via cmark-gfm).
    Reimplemented reference storage as follows:
    1. New references are always inserted at the end of a linked list. This
    is an O(1) operation, and does not check whether an existing (duplicate)
    reference with the same label already exists in the document.
    2. Upon the first call to `cmark_reference_lookup` (when it is expected
    that no further references will be added to the reference map), the
    linked list of references is written into a fixed-size array.
    3. The fixed size array can then be efficiently sorted in-place in O(n
    log n). This operation only happens once. We perform this sort in a
    _stable_ manner to ensure that the earliest link reference in the
    document always has preference, as the spec dictates. To accomplish
    this, every reference is tagged with a generation number when initially
    inserted in the linked list.
    4. The sorted array is then compacted in O(n). Since it was sorted in a
    stable way, the first reference for each label is preserved and the
    duplicates are removed, matching the spec.
    5. We can now simply perform a binary search for the current
    `cmark_reference_lookup` query in O(log n). Any further lookup calls
    will also be O(log n), since the sorted references table only needs to
    be generated once.
    The resulting implementation is notably simple (as it uses standard
    library builtins `qsort` and `bsearch`), whilst performing better than
    the fixed size hash table in documents that have a high number of
    references and never becoming pathological regardless of the input.
  * Comment out unused function `cmark_strbuf_cstr` in `buffer.h`.
  * Re-add `--safe` command-line option as a no-op (#344), for backwards
    compatibility.
  * Update to Unicode 13.0
  * Generate and install cmake-config file (Reinhold Gschweicher).
    Add full cmake support. The project can either be used with
    `add_subdirectory` or be installed into the system (or some other
    directory) and be found with `find_package(cmark)`. In both cases the
    cmake target `cmark::cmark` and/or `cmark::cmark_static` is all that
    is needed to be linked.  Previously the `cmarkConfig.cmake` file
    was generated, but not installed.  As additional bonus of generation
    by cmake we get a generated `cmake-config-version.cmake` file for
    `find_package()` to search for the same major version.
    The generated config file is position independent, allowing the
    installed directory to be copied or moved and still work.
    The following four files are generated and installed:
    `lib/cmake/cmark/cmark-config.cmake`,
    `lib/cmake/cmark/cmark-config-version.cmake`,
    `lib/cmake/cmark/cmark-targets.cmake`,
    `lib/cmake/cmark/cmark-targets-release.cmake`.
  * Adjust the MinGW paths for MinGW64 (Daniil Baturin).
  * Fix CMake generator expression checking for MSVC (Nick Wellnhofer).
  * Fix `-Wconst-qual` warning (Saleem Abdulrasool).  This enables building
    with `/Zc:strictString` with MSVC as well.
  * Improve and modernize cmake build (Saleem Abdulrasool).
    + Build: add exports targets for build tree usage (#307).
    + Uuse target properties for include paths.
    + Remove the unnecessary execute permission on CMakeLists.txt.
    + Reduce property computation in CMake.
    + Use `CMAKE_INCLUDE_CURRENT_DIRECTORY`.
    + Improve man page installation.
    + Only include `GNUInstallDirs` once.
    + Replace `add_compile_definitions` with `add_compile_options`
      since the former was introduced in 3.12 (#321).
    + Cleanup CMake (#319).
    + Inline a variable.
    + Use `LINKER_LANGUAGE` property for C++ runtime.
    + Use CMake to control C standard.
    + Use the correct variable.
    + Loosen the compiler check
    + Hoist shared flags to top-level CMakeLists
    + Remove duplicated flags.
    + Use `add_compile_options` rather than modify `CMAKE_C_FLAGS`.
    + Hoist sanitizer flags to global state.
    + Hoist `-fvisibilty` flags to top-level.
    + Hoist the debug flag handling.
    + Hoist the profile flag handling.
    + Remove incorrect variable handling.
    + Remove unused CMake includes.
  * Remove "-rdynamic" flag for static builds (#300, Eric Pruitt).
  * Fixed installation on other than Ubuntu GNU/Linux distributions
    (Vitaly Zaitsev).
  * Link executable with static or shared library (Nick Wellnhofer).
    If `CMARK_STATIC` is on (default), link the executable with the static
    library. This produces exactly the same result as compiling the library
    sources again and linking with the object files.
    If `CMARK_STATIC` is off, link the executable with the shared library.
    This wasn't supported before and should be the preferred way to
    package cmark on Linux distros.
    Building only a shared library and a statically linked executable
    isn't supported anymore but this doesn't seem useful.
  * Reintroduce version check for MSVC /TP flag (Nick Wellnhofer).
    The flag is only required for old MSVC versions.
  * normalize.py: use `html.escape` instead of `cgi.escape` (#313).
  * Fix pathological_tests.py on Windows (Nick Wellnhofer).
    When using multiprocessing on Windows, the main program must be
    guarded with a `__name__` check.
  * Remove useless `__name__` check in test scripts (Nick Wellnhofer).
  * Add CIFuzz (Leo Neat).
  * cmark.1 - Document --unsafe instead of --safe (#332).
  * cmark.1: remove docs for `--normalize` which no longer exists (#332).
  * Add lint target to Makefile.
  * Add uninstall target to Makefile.
  * Update benchmarks (#338).
  * Fix typo in documentation (Tim Gates).
  * Increase timeout for pathological tests to avoid CI failure.
  * Update the Racket wrapper with the safe -> unsafe flag change (Eli
    Barzilay).

[0.29.0]

  * Update spec to 0.29.
  * Make rendering safe by default (#239, #273).
    Adds `CMARK_OPT_UNSAFE` and make `CMARK_OPT_SAFE` a no-op (for API
    compatibility).  The new default behavior is to suppress raw HTML and
    potentially dangerous links.  The `CMARK_OPT_UNSAFE` option has to be set
    explicitly to prevent this.
    **NOTE:** This change will require modifications in bindings for cmark
    and in most libraries and programs that use cmark.
    Borrows heavily from @kivikakk's patch in github/cmark-gfm#123.
  * Add sourcepos info for inlines (Yuki Izumi).
  * Disallow more than 32 nested balanced parens in a link (Yuki Izumi).
  * Resolve link references before creating setext header.
    A setext header line after a link reference should not
    create a header, according to the spec.
  * commonmark renderer: improve escaping.
    URL-escape special characters when escape mode is URL, and not otherwise.
    Entity-escape control characters (< 0x20) in non-literal escape modes.
  * render:  only emit actual newline when escape mode is LITERAL.
    For markdown content, e.g., in other contexts we want some
    kind of escaping, not a literal newline.
  * Update code span normalization to conform with spec change.
  * Allow empty `<>` link destination in reference link.
  * Remove leftover includes of `memory.h` (#290).
  * A link destination can't start with `<` unless it is
    an angle-bracket link that also ends with `>` (#289).
    (If your URL really starts with `<`, URL-escape it.)
  * Allow internal delimiter runs to match if both have lengths that are
    multiples of 3.  See commonmark/commonmark#528.
  * Include `references.h` in `parser.h` (#287).
  * Fix `[link](<foo\>)`.
  * Use hand-rolled scanner for thematic break (see #284).
    Keep track of the last position where a thematic break
    failed to match on a line, to avoid rescanning unnecessarily.
  * Rename `ends_with_blank_line` with `S_` prefix.
  * Add `CMARK_NODE__LAST_LINE_CHECKED` flag (#284).
    Use this to avoid unnecessary recursion in `ends_with_blank_line`.
  * In `ends_with_blank_line`, call `S_set_last_line_blank`
    to avoid unnecessary repetition (#284).  Once we settle whether a list
    item ends in a blank line, we don't need to revisit this in considering
    parent list items.
  * Disallow unescaped `(` in parenthesized link title.
  * Copy line/col info straight from opener/closer (Ashe Connor).
    We can't rely on anything in `subj` since it's been modified while parsing
    the subject and could represent line info from a future line.  This is
    simple and works.
  * `render.c`: reset `last_breakable` after cr.  Fixes jgm/pandoc#5033.
  * Fix a typo in `houdini_href_e.c` (Felix Yan).
  * commonmark writer: use `~~~` fences if info string contains backtick.
    This is needed for round-trip tests.
  * Update scanners for new info string rules.
  * Add XSLT stylesheet to convert cmark XML back to Commonmark
    (Nick Wellnhofer, #264).  Initial version of an XSLT stylesheet that
    converts the XML format produced by `cmark -t xml` back to Commonmark.
  * Check for whitespace before reference title (#263).
  * Bump CMake to version 3 (Jonathan Müller).
  * Build: Remove deprecated call to `add_compiler_export_flags()`
    (Jonathan Müller).  It is deprecated in CMake 3.0, the replacement is to
    set the `CXX_VISIBILITY_PRESET` (or in our case `C_VISIBILITY_PRESET`) and
    `VISIBILITY_INLINES_HIDDEN` properties of the target.  We're already
    setting them by setting the CMake variables anyway, so the call can be
    removed.
  * Build: only attempt to install MSVC system libraries on Windows
    (Saleem Abdulrasool).  Newer versions of CMake attempt to query the system
    for information about the VS 2017 installation.  Unfortunately, this query
    fails on non-Windows systems when cross-compiling:
    `cmake_host_system_information does not recognize <key> VS_15_DIR`.
    CMake will not find these system libraries on non-Windows hosts anyways,
    and we were silencing the warnings, so simply omit the installation when
    cross-compiling to Windows.
  * Simplify code normalization, in line with spec change.
  * Implement code span spec changes.  These affect both parsing and writing
    commonmark.
  * Add link parsing corner cases to regressions (Ashe Connor).
  * Add `xml:space="preserve"` in XML output when appropriate
    (Nguyễn Thái Ngọc Duy).
    (For text, code, code_block, html_inline and html_block tags.)
  * Removed meta from list of block tags.  Added regression test.
    See commonmark/CommonMark#527.
  * `entity_tests.py` - omit noisy success output.
  * `pathological_tests.py`: make tests run faster.
    Commented out the (already ignored) "many references" test, which
    times out.  Reduced the iterations for a couple other tests.
  * `pathological_tests.py`: added test for deeply nested lists.
  * Optimize `S_find_first_nonspace`.  We were needlessly redoing things we'd
    already done.  Now we skip the work if the first nonspace is greater than
    the current offset.  This fixes pathological slowdown with deeply nested
    lists (#255).  For N = 3000, the time goes from over 17s to about 0.7s.
    Thanks to Martin Mitas for diagnosing the problem.
  * Allow spaces in link destination delimited with pointy brackets.
  * Adjust max length of decimal/numeric entities.
    See commonmark/CommonMark#487.
  * Fix inline raw HTML parsing.
    This fixes a recently added failing spec test case.  Previously spaces
    were being allowed in unquoted attribute values; no we forbid them.
  * Don't allow list markers to be indented >= 4 spaces.
    See commonmark/CommonMark#497.
  * Check for empty buffer when rendering (Phil Turnbull).
    For empty documents, `->size` is zero so
    `renderer.buffer->ptr[renderer.buffer->size - 1]` will cause an
    out-of-bounds read. Empty buffers always point to the global
    `cmark_strbuf__initbuf` buffer so we read `cmark_strbuf__initbuf[-1]`.
  * Also run API tests with `CMARK_SHARED=OFF` (Nick Wellnhofer).
  * Rename roundtrip and entity tests (Nick Wellnhofer).
    Rename the tests to reflect that they use the library, not the
    executable.
  * Generate export header for static-only build (#247, Nick Wellnhofer).
  * Fuzz width parameter too (Phil Turnbull).  Allow the `width` parameter to
    be generated too so we get better fuzz-coverage.
  * Don't discard empty fuzz test-cases (Phil Turnbull).  We currently discard
    fuzz test-cases that are empty but empty inputs are valid markdown. This
    improves the fuzzing coverage slightly.
  * Fixed exit code for pathological tests.
  * Add allowed failures to `pathological_tests.py`.
    This allows us to include tests that we don't yet know how to pass.
  * Add timeout to `pathological_tests.py`.
    Tests must complete in 8 seconds or are errors.
  * Add more pathological tests (Martin Mitas).
    These tests target the issues #214, #218, #220.
  * Use pledge(2) on OpenBSD (Ashe Connor).
  * Update the Racket wrapper (Eli Barzilay).
  * Makefile: For afl target, don't build tests.

[0.28.3]

  * Include GNUInstallDirs in src/CMakeLists.txt (Nick Wellnhofer, #240).
    This fixes build problems on some cmake versions (#241).

[0.28.2]

  * Fixed regression in install dest for static library (#238).
    Due to a mistake, 0.28.1 installed libcmark.a into include/.

[0.28.1]

  * `--smart`: open quote can never occur right after `]` or `)` (#227).
  * Fix quadratic behavior in `finalize` (Vicent Marti).
  * Don't use `CMAKE_INSTALL_LIBDIR` to create `libcmark.pc` (#236).
    This wasn't getting set in processing `libcmark.pc.in`, and we
    were getting the wrong entry in `libcmark.pc`.
    The new approach sets an internal `libdir` variable to
    `lib${LIB_SUFFIX}`.  This variable is used both to set the
    install destination and in the libcmark.pc.in template.
  * Update README.md, replace `make astyle` with `make format`
    (Nguyễn Thái Ngọc Duy).

[0.28.0]

  * Update spec.
  * Use unsigned integer when shifting (Phil Turnbull).
    Avoids a UBSAN warning which can be triggered when handling a
    long sequence of backticks.
  * Avoid memcpy'ing NULL pointers (Phil Turnbull).
    Avoids a UBSAN warning when link title is empty string.
    The length of the memcpy is zero so the NULL pointer is not
    dereferenced but it is still undefined behaviour.
  * DeMorgan simplification of some tests in emphasis parser.
    This also brings the code into closer alignment with the wording
    of the spec (see jgm/CommonMark#467).
  * Fixed undefined shift in commonmark writer (#211).
    Found by google/oss-fuzz:
    <https://oss-fuzz.com/v2/testcase-detail/4686992824598528>.
  * latex writer:  fix memory overflow (#210).
    We got an array overflow in enumerated lists nested more than
    10 deep with start number =/= 1.
    This commit also ensures that we don't try to set `enum_` counters
    that aren't defined by LaTeX (generally up to enumv).
    Found by google/oss-fuzz:
    <https://oss-fuzz.com/v2/testcase-detail/5546760854306816>.
  * Check for NULL pointer in get_link_type (Phil Turnbull).
    `echo '[](xx:)' | ./build/src/cmark -t latex` gave a
    segfault.
  * Move fuzzing dictionary into single file (Phil Turnbull).
    This allows AFL and libFuzzer to use the same dictionary
  * Reset bytes after UTF8 proc (Yuki Izumi, #206).
  * Don't scan past an EOL (Yuki Izumi).
    The existing negated character classes (`[^…]`) are careful to
    always include` \x00` in the characters excluded, but these `.`
    catch-alls can scan right past the terminating NUL placed
    at the end of the buffer by `_scan_at`.  As such, buffer
    overruns can occur.  Also, don't scan past a newline in HTML
    block end scanners.
  * Document cases where `get_` functions return `NULL` (#155).
    E.g. `cmark_node_get_url` on a non-link or image.
  * Properly handle backslashes in link destinations (#192).
    Only ascii punctuation characters are escapable, per the spec.
  * Fixed `cmark_node_get_list_start` to return 0 for bullet lists,
    as documented (#202).
  * Use `CMARK_NO_DELIM` for bullet lists (#201).
  * Fixed code for freeing delimiter stack (#189).
  * Removed abort outside of conditional (typo).
  * Removed coercion in error message when aborting from buffer.
  * Print message to stderr when we abort due to memory demands (#188).
  * `libcmark.pc`: use `CMAKE_INSTALL_LIBDIR` (#185, Jens Petersen).
    Needed for multilib distros like Fedora.
  * Fixed buffer overflow error in `S_parser_feed` (#184).
    The overflow could occur in the following condition:
    the buffer ends with `\r` and the next memory address
    contains `\n`.
  * Update emphasis parsing for spec change.
    Strong now goes inside Emph rather than the reverse,
    when both scopes are possible.  The code is much simpler.
    This also avoids a spec inconsistency that cmark had previously:
    `***hi***` became Strong (Emph "hi")) but
    `***hi****` became Emph (Strong "hi")) "*"
  * Fixes for the LaTeX renderer (#182, Doeme)
    + Don't double-output the link in latex-rendering.
    + Prevent ligatures in dashes sensibly when rendering latex.
      `\-` is a hyphenation, so it doesn't get displayed at all.
  * Added a test for NULL when freeing `subj->last_delim`.
  * Cleaned up setting of lower bounds for openers.
    We now use a much smaller array.
  * Fix #178, quadratic parsing bug.  Add pathological test.
  * Slight improvement of clarity of logic in emph matching.
  * Fix "multiple of 3" determination in emph/strong parsing.
    We need to store the length of the original delimiter run,
    instead of using the length of the remaining delimiters
    after some have been subtracted.  Test case:
    `a***b* c*`.  Thanks to Raph Levin for reporting.
  * Correctly initialize chunk in S_process_line (Nick Wellnhofer, #170).
    The `alloc` member wasn't initialized.  This also allows to add an
    assertion in `chunk_rtrim` which doesn't work for alloced chunks.
  * Added 'make newbench'.
  * `scanners.c` generated with re2c 0.16 (68K smaller!).
  * `scanners.re` - fixed warnings; use `*` for fallback.
  * Fixed some warnings in `scanners.re`.
  * Update CaseFolding to latest (Kevin Wojniak, #168).
  * Allow balanced nested parens in link destinations (Yuki Izumi, #166)
  * Allocate enough bytes for backticks array.
  * Inlines: Ensure that the delimiter stack is freed in subject.
  * Fixed pathological cases with backtick code spans:

    - Removed recursion in scan_to_closing_backticks
    - Added an array of pointers to potential backtick closers
      to subject
    - This array is used to avoid traversing the subject again
      when we've already seen all the potential backtick closers.
    - Added a max bound of 1000 for backtick code span delimiters.
    - This helps with pathological cases like:

            x
            x `
            x ``
            x ```
            x ````
            ...

    - Added pathological test case.

    Thanks to Martin Mitáš for identifying the problem and for
    discussion of solutions.
  * Remove redundant cmake_minimum_required (#163, @kainjow).
  * Make shared and static libraries optional (Azamat H. Hackimov).
    Now you can enable/disable compilation and installation targets for
    shared and static libraries via `-DCMARK_SHARED=ON/OFF` and
    `-DCMARK_STATIC=ON/OFF`.
  * Added support for built-in `${LIB_SUFFIX}` feature (Azamat H.
    Hackimov).  Replaced `${LIB_INSTALL_DIR}` option with built-in
    `${LIB_SUFFIX}` for installing for 32/64-bit systems. Normally,
    CMake will set `${LIB_SUFFIX}` automatically for required enviroment.
    If you have any issues with it, you can override this option with
    `-DLIB_SUFFIX=64` or `-DLIB_SUFFIX=""` during configuration.
  * Add Makefile target and harness to fuzz with libFuzzer (Phil Turnbull).
    This can be run locally with `make libFuzzer` but the harness will be
    integrated into oss-fuzz for large-scale fuzzing.
  * Advertise `--validate-utf8` in usage information
    (Nguyễn Thái Ngọc Duy).
  * Makefile: use warnings with re2c.
  * README: Add link to Python wrapper, prettify languages list
    (Pavlo Kapyshin).
  * README: Add link to cmark-scala (Tim Nieradzik, #196)

[0.27.1]

  * Set policy for CMP0063 to avoid a warning (#162).
    Put set_policy under cmake version test.
    Otherwise we get errors in older versions of cmake.
  * Use VERSION_GREATER to clean up cmake version test.
  * Improve afl target.  Use afl-clang by default.  Set default for path.

[0.27.0]

  * Update spec to 0.27.
  * Fix warnings building with MSVC on Windows (#165, Hugh Bellamy).
  * Fix `CMAKE_C_VISIBILITY_PRESET` for cmake versions greater than 1.8
    (e.g. 3.6.2) (#162, Hugh Bellamy).  This lets us build swift-cmark
    on Windows, using clang-cl.
  * Fix for non-matching entities (#161, Yuki Izumi).
  * Modified `print_delimiters` (commented out) so it compiles again.
  * `make format`: don't change order of includes.
  * Changed logic for null/eol checks (#160).
    + only check once for "not at end of line"
    + check for null before we check for newline characters (the
      previous patch would fail for NULL + CR)
  * Fix by not advancing past both `\0` and `\n` (Yuki Izumi).
  * Add test for NUL-LF sequence (Yuki Izumi).
  * Fix memory leak in list parsing (Yuki Izumi).
  * Use `cmark_mem` to free where used to alloc (Yuki Izumi).
  * Allow a shortcut link before a `(` (jgm/CommonMark#427).
  * Allow tabs after setext header line (jgm/commonmark.js#109).
  * Don't let URI schemes start with spaces.
  * Fixed h2..h6 HTML blocks (jgm/CommonMark#430).  Added regression test.
  * Autolink scheme can contain digits (Gábor Csárdi).
  * Fix nullary function declarations in cmark.h (Nick Wellnhofer).
    Fixes strict prototypes warnings.
  * COPYING: Update file name and remove duplicate section and
    (Peter Eisentraut).
  * Fix typo (Pavlo Kapyshin).

[0.26.1]

  * Removed unnecessary typedef that caused build failure on
    some platforms.
  * Use `$(MAKE)` in Makefile instead of hardcoded `make` (#146,
    Tobias Kortkamp).

[0.26.0]

  * Implement spec changes for list items:
    - Empty list items cannot interrupt paragraphs.
    - Ordered lists cannot interrupt paragraphs unless
      they start with 1.
    - Removed "two blank lines break out of a list" feature.
  * Fix sourcepos for blockquotes (#142).
  * Fix sourcepos for atx headers (#141).
  * Fix ATX headers and thematic breaks to allow tabs as well as spaces.
  * Fix `chunk_set_cstr` with suffix of current string (#139,
    Nick Wellnhofer).  It's possible that `cmark_chunk_set_cstr` is called
    with a substring (suffix) of the current string. Delay freeing of the
    chunk content to handle this case correctly.
  * Export targets on installation (Jonathan Müller).
    This allows using them in other cmake projects.
  * Fix cmake warning about CMP0048 (Jonathan Müller).
  * commonmark renderer:  Ensure we don't have a blank line
    before a code block when it's the first thing in a list
    item.
  * Change parsing of strong/emph in response to spec changes.
    `process_emphasis` now gets better results in corner cases.
    The change is this:  when considering matches between an interior
    delimiter run (one that can open and can close) and another delimiter
    run, we require that the sum of the lengths of the two delimiter
    runs mod 3 is not 0.
  * Ported Robin Stocker's changes to link parsing in jgm/commonmark#101.
    This uses a separate stack for brackets, instead of putting them on the
    delimiter stack.  This avoids the need for looking through the delimiter
    stack for the next bracket.
  * `cmark_reference_lookup`: Return NULL if reference is null string.
  * Fix character type detection in `commonmark.c` (Nick Wellnhofer).
    Fixes test failures on Windows and undefined behavior.
    - Implement `cmark_isalpha`.
    - Check for ASCII character before implicit cast to char.
    - Use internal ctype functions in `commonmark.c`.
  * Better documentation of memory-freeing responsibilities.
    in `cmark.h` and its man page (#124).
  * Use library functions to insert nodes in emphasis/link processing.
    Previously we did this manually, which introduces many
    places where errors can creep in.
  * Correctly handle list marker followed only by spaces.
    Previously when a list marker was followed only by spaces,
    cmark expected the following content to be indented by
    the same number of spaces.  But in this case we should
    treat the line just like a blank line and set list padding
    accordingly.
  * Fixed a number of issues relating to line wrapping.
    - Extend `CMARK_OPT_NOBREAKS` to all renderers and add `--nobreaks`.
    - Do not autowrap, regardless of width parameter, if
      `CMARK_OPT_NOBREAKS` is set.
    - Fixed `CMARK_OPT_HARDBREAKS` for LaTeX and man renderers.
    - Ensure that no auto-wrapping occurs if
      `CMARK_OPT_NOBREAKS` is enabled, or if output is CommonMark and
      `CMARK_OPT_HARDBREAKS` is enabled.
  * Set stdin to binary mode on Windows (Nick Wellnhofer, #113).
    This fixes EOLs when reading from stdin.
  * Add library option to render softbreaks as spaces (Pavlo
    Kapyshin).  Note that the `NOBREAKS` option is HTML-only
  * renderer:  `no_linebreaks` instead of `no_wrap`.
    We generally want this option to prohibit any breaking
    in things like headers (not just wraps, but softbreaks).
  * Coerce `realurllen` to `int`.  This is an alternate solution for pull
    request #132, which introduced a new warning on the comparison
    (Benedict Cohen).
  * Remove unused variable `link_text` (Mathiew Duponchelle).
  * Improved safety checks in buffer (Vicent Marti).
  * Add new interface allowing specification of custom memory allocator
    for nodes (Vicent Marti).  Added `cmark_node_new_with_mem`,
    `cmark_parser_new_with_mem`, `cmark_mem` to API.
  * Reduce storage size for nodes by using bit flags instead of
    separate booleans (Vicent Marti).
  * config: Add `SSIZE_T` compat for Win32 (Vicent Marti).
  * cmake: Global handler for OOM situations (Vicent Marti).
  * Add tests for memory exhaustion (Vicent Marti).
  * Document in man page and public header that one should use the same
    memory allocator for every node in a tree.
  * Fix ctypes in Python FFI calls (Nick Wellnhofer).  This didn't cause
    problems so far because all types are 32-bit on 32-bit systems and
    arguments are passed in registers on x86-64.  The wrong types could cause
    crashes on other platforms, though.
  * Remove spurious failures in roundtrip tests.  In the commonmark writer we
    separate lists, and lists and indented code, using a dummy HTML comment.
    So in evaluating the round-trip tests, we now strip out
    these comments.  We also normalize HTML to avoid issues
    having to do with line breaks.
  * Add 2016 to copyright (Kevin Burke).
  * Added `to_commonmark` in `test/cmark.py` (for round-trip tests).
  * `spec_test.py` - parameterize `do_test` with converter.
  * `spec_tests.py`: exit code is now sum of failures and errors.
    This ensures that a failing exit code will be given when
    there are errors, not just with failures.
  * Fixed round trip tests.  Previously they actually ran
    `cmark` instead of the round-trip version, since there was a bug in
    setting the ROUNDTRIP variable (#131).
  * Added new `roundtrip_tests.py`.  This replaces the old use of simple shell
    scripts.  It is much faster, and more flexible.  (We will be able
    to do custom normalization and skip certain tests.)
  * Fix tests under MinGW (Nick Wellnhofer).
  * Fix leak in `api_test` (Mathieu Duponchelle).
  * Makefile: have leakcheck stop on first error instead of going through
    all the formats and options and probably getting the same output.
  * Add regression tests (Nick Wellnhofer).

[0.25.2]

  * Open files in binary mode (#113, Nick Wellnhofer).  Now that cmark
    supports different line endings, files must be openend in binary mode
    on Windows.
  * Reset `partially_consumed_tab` on every new line (#114, Nick Wellnhofer).
  * Handle buffer split across a CRLF line ending (#117).  Adds an internal
    field to the parser struct to keep track of `last_buffer_ended_with_cr`.
    Added test.

[0.25.1]

  * Release with no code changes.  cmark version was mistakenly set to
    0.25.1 in the 0.25.0 release (#112), so this release just
    ensures that this will cause no confusion later.

[0.25.0]

  * Fixed tabs in indentation (#101).  This patch fixes S_advance_offset
    so that it doesn't gobble a tab character when advancing less than the
    width of a tab.
  * Added partially_consumed_tab to parser.  This keeps track of when we
    have gotten partway through a tab when consuming initial indentation.
  * Simplified add_line (only need parser parameter).
  * Properly handle partially consumed tab.  E.g. in

        - foo

         <TAB><TAB>bar

    we should consume two spaces from the second tab, including two spaces
    in the code block.
  * Properly handle tabs with blockquotes and fenced blocks.
  * Fixed handling of tabs in lists.
  * Clarified logic in S_advance_offset.
  * Use an assertion to check for in-range html_block_type.
    It's a programming error if the type is out of range.
  * Refactored S_processLines to make the logic easier to
    understand, and added documentation (Mathieu Duponchelle).
  * Removed unnecessary check for empty string_content.
  * Factored out contains_inlines.
  * Moved the cmake minimum version to top line of CMakeLists.txt
    (tinysun212).
  * Fix ctype(3) usage on NetBSD (Kamil Rytarowski).  We need to cast value
    passed to isspace(3) to unsigned char to explicitly prevent possibly
    undefined behavior.
  * Compile in plain C mode with MSVC 12.0 or newer (Nick Wellnhofer).
    Under MSVC, we used to compile in C++ mode to get some C99 features
    like mixing declarations and code. With newer MSVC versions, it's
    possible to build in plain C mode.
  * Switched from "inline" to "CMARK_INLINE" (Nick Wellnhofer).
    Newer MSVC versions support enough of C99 to be able to compile cmark
    in plain C mode. Only the "inline" keyword is still unsupported.
    We have to use "__inline" instead.
  * Added include guards to config.h
  * config.h.in - added compatibility snprintf, vsnprintf for MSVC.
  * Replaced sprintf with snprintf (Marco Benelli).
  * config.h: include stdio.h for _vscprintf etc.
  * Include starg.h when needed in config.h.
  * Removed an unnecessary C99-ism in buffer.c.  This helps compiling on
    systems like luarocks that don't have all the cmake configuration
    goodness (thanks to carlmartus).
  * Don't use variable length arrays (Nick Wellnhofer).
    They're not supported by MSVC.
  * Test with multiple MSVC versions under Appveyor (Nick Wellnhofer).
  * Fix installation dir of man-pages on NetBSD (Kamil Rytarowski).
  * Fixed typo in cmark.h comments (Chris Eidhof).
  * Clarify in man page that cmark_node_free frees a node's children too.
  * Fixed documentation of --width in man page.
  * Require re2c >= 1.14.2 (#102).
  * Generated scanners.c with more recent re2c.

[0.24.1]

  * Commonmark renderer:
    + Use HTML comment, not two blank lines, to separate a list
      item from a following code block or list.  This makes the
      output more portable, since the "two blank lines" rule is
      unique to CommonMark.  Also, it allows us to break out of
      a sublist without breaking out of all levels of nesting.
    + `is_autolink` - handle case where link has no children,
      which previously caused a segfault.
    + Use 4-space indent for bullet lists, for increased portability.
    + Use 2-space + newline for line break for increased portability (#90).
    + Improved punctuation escaping.  Previously all `)` and
      `.` characters after digits were escaped; now they are
      only escaped if they are genuinely in a position where
      they'd cause a list item.  This is achieved by changes in
      `render.c`: (a) `renderer->begin_content` is only set to
      false after a string of digits at the beginning of the
      line, and (b) we never break a line before a digit.
      Also, `begin_content` is properly initialized to true.
  * Handle NULL root in `consolidate_text_nodes`.

[0.24.0]

  * [API change] Added `cmark_node_replace(oldnode, newnode)`.
  * Updated spec.txt to 0.24.
  * Fixed edge case with escaped parens in link destination (#97).
    This was also checked against the #82 case with asan.
  * Removed unnecessary check for `fenced` in `cmark_render_html`.
    It's sufficient to check that the info string is empty.
    Indeed, those who use the API may well create a code block
    with an info string without explicitly setting `fenced`.
  * Updated format of `test/smart_punct.txt`.
  * Updated `test/spec.txt`, `test/smart_punct.txt`, and
    `spec_tests.py` to new format.
  * Fixed `get_containing_block` logic in `src/commonmark.c`.
    This did not allow for the possibility that a node might have no
    containing block, causing the commonmark renderer to segfault if
    passed an inline node with no block parent.
  * Fixed string representations of `CUSTOM_BLOCK`,
    `CUSTOM_INLINE`.  The old versions `raw_inline` and
    `raw_block` were being used, and this led to incorrect xml output.
  * Use default opts in python sample wrapper.
  * Allow multiline setext header content, as per spec.
  * Don't allow spaces in link destinations, even with pointy brackets.
    Conforms to latest change in spec.
  * Updated `scheme` scanner according to spec change.  We no longer use
    a whitelist of valid schemes.
  * Allow any kind of nodes as children of `CUSTOM_BLOCK` (#96).
  * `cmark.h`: moved typedefs for iterator into iterator section.
    This just moves some code around so it makes more sense
    to read, and in the man page.
  * Fixed `make_man_page.py` so it includes typedefs again.

[0.23.0]

  * [API change] Added `CUSTOM_BLOCK` and `CUSTOM_INLINE` node types.
    They are never generated by the parser, and do not correspond
    to CommonMark elements.  They are designed to be inserted by
    filters that postprocess the AST.  For example, a filter might
    convert specially marked code blocks to svg diagrams in HTML
    and tikz diagrams in LaTeX, passing these through to the renderer
    as a `CUSTOM_BLOCK`.  These nodes can have children, but they
    also have literal text to be printed by the renderer "on enter"
    and "on exit." Added `cmark_node_get_on_enter`,
    `cmark_node_set_on_enter`, `cmark_node_get_on_exit`,
    `cmark_node_set_on_exit` to API.
  * [API change] Rename `NODE_HTML` -> `NODE_HTML_BLOCK`,
    `NODE_INLINE_HTML` -> `NODE_HTML_INLINE`.  Define aliases
    so the old names still work, for backwards compatibility.
  * [API change] Rename `CMARK_NODE_HEADER` -> `CMARK_NODE_HEADING`.
    Note that for backwards compatibility, we have defined aliases:
    `CMARK_NODE_HEADER` = `CMARK_NODE_HEADING`,
    `cmark_node_get_header_level` = `cmark_node_get_heading_level`, and
    `cmark_node_set_header_level` = `cmark_node_set_heading_level`.
  * [API change] Rename `CMARK_NODE_HRULE` -> `CMARK_NODE_THEMATIC_BREAK`.
    Defined the former as the latter for backwards compatibility.
  * Don't allow space between link text and link label in a reference link
    (spec change).
  * Separate parsing and rendering opts in `cmark.h` (#88).
    This change also changes some of these constants' numerical values,
    but nothing should change in the API if you use the constants
    themselves.  It should now be clear in the man page which
    options affect parsing and which affect rendering.
  * xml renderer - Added xmlns attribute to document node (jgm/CommonMark#87).
  * Commonmark renderer:  ensure html blocks surrounded by blanks.
    Otherwise we get failures of roundtrip tests.
  * Commonmark renderer: ensure that literal characters get escaped
    when they're at the beginning of a block, e.g.  `> \- foo`.
  * LaTeX renderer - better handling of internal links.
    Now we render `[foo](#bar)` as `\protect\hyperlink{bar}{foo}`.
  * Check for NULL pointer in _scan_at (#81).
  * `Makefile.nmake`:  be more robust when cmake is missing.  Previously,
    when cmake was missing, the build dir would be created anyway, and
    subsequent attempts (even with cmake) would fail, because cmake would
    not be run.  Depending on `build/CMakeFiles` is more robust -- this won't
    be created unless cmake is run.  Partially addresses #85.
  * Fixed DOCTYPE in xml output.
  * commonmark.c: fix `size_t` to `int`.  This fixes an MSVC warning
   "conversion from 'size_t' to 'int', possible loss of data" (Kevin Wojniak).
  * Correct string length in `cmark_parse_document` example (Lee Jeffery).
  * Fix non-ASCII end-of-line character check (andyuhnak).
  * Fix "declaration shadows a local variable" (Kevin Wojniak).
  * Install static library (jgm/CommonMark#381).
  * Fix warnings about dropping const qualifier (Kevin Wojniak).
  * Use full (unabbreviated) versions of constants (`CMARK_...`).
  * Removed outdated targets from Makefile.
  * Removed need for sudo in `make bench`.
  * Improved benchmark.  Use longer test, since `time` has limited resolution.
  * Removed `bench.h` and timing calls in `main.c`.
  * Updated API docs; getters return empty strings if not set
    rather than NULL, as previously documented.
  * Added api_tests for custom nodes.
  * Made roundtrip test part of the test suite run by cmake.
  * Regenerate `scanners.c` using re2c 0.15.3.
  * Adjusted scanner for link url.  This fixes a heap buffer overflow (#82).
  * Added version number (1.0) to XML namespace.  We don't guarantee
    stability in this until 1.0 is actually released, however.
  * Removed obsolete `TIMER` macro.
  * Make `LIB_INSTALL_DIR` configurable (Mathieu Bridon, #79).
  * Removed out-of-date luajit wrapper.
  * Use `input`, not `parser->curline` to determine last line length.
  * Small optimizations in `_scan_at`.
  * Replaced hard-coded 4 with `TAB_STOP`.
  * Have `make format` reformat api tests as well.
  * Added api tests for man, latex, commonmark, and xml renderers (#51).
  * render.c:  added `begin_content` field.  This is like `begin_line` except
    that it doesn't trigger production of the prefix.  So it can be set
    after an initial prefix (say `> `) is printed by the renderer, and
    consulted in determining whether to escape content that has a special
    meaning at the beginning of a line.  Used in the commonmark renderer.
  * Python 3.5 compatibility: don't require HTMLParseError (Zhiming Wang).
    HTMLParseError was removed in Python 3.5. Since it could never be thrown
    in Python 3.5+, we simply define a placeholder when HTMLParseError
    cannot be imported.
  * Set `convert_charrefs=False` in `normalize.py` (#83).  This defeats the
    new default as of python 3.5, and allows the script to work with python
    3.5.

[0.22.0]

  * Removed `pre` from blocktags scanner. `pre` is handled separately
    in rule 1 and needn't be handled in rule 6.
  * Added `iframe` to list of blocktags, as per spec change.
  * Fixed bug with `HRULE` after blank line. This previously caused cmark
    to break out of a list, thinking it had two consecutive blanks.
  * Check for empty string before trying to look at line ending.
  * Make sure every line fed to `S_process_line` ends with `\n` (#72).
    So `S_process_line` sees only unix style line endings. Ultimately we
    probably want a better solution, allowing the line ending style of
    the input file to be preserved. This solution forces output with newlines.
  * Improved `cmark_strbuf_normalize_whitespace` (#73). Now all characters
    that satisfy `cmark_isspace` are recognized as whitespace. Previously
    `\r` and `\t` (and others) weren't included.
  * Treat line ending with EOF as ending with newline (#71).
  * Fixed `--hardbreaks` with `\r\n` line breaks (#68).
  * Disallow list item starting with multiple blank lines (jgm/CommonMark#332).
  * Allow tabs before closing `#`s in ATX header
  * Removed `cmark_strbuf_printf` and `cmark_strbuf_vprintf`.
    These are no longer needed, and cause complications for MSVC.
    Also removed `HAVE_VA_COPY` and `HAVE_C99_SNPRINTF` feature tests.
  * Added option to disable tests (Kevin Wojniak).
  * Added `CMARK_INLINE` macro.
  * Removed need to disable MSVC warnings 4267, 4244, 4800
    (Kevin Wojniak).
  * Fixed MSVC inline errors when cmark is included in sources that
    don't have the same set of disabled warnings (Kevin Wojniak).
  * Fix `FileNotFoundError` errors on tests when cmark is built from
    another project via `add_subdirectory()` (Kevin Wojniak).
  * Prefix `utf8proc` functions to avoid conflict with existing library
    (Kevin Wojniak).
  * Avoid name clash between Windows `.pdb` files (Nick Wellnhofer).
  * Improved `smart_punct.txt` (see jgm/commonmark.js#61).
  * Set `POSITION_INDEPENDENT_CODE` `ON` for static library (see #39).
  * `make bench`: allow overriding `BENCHFILE`. Previously if you did
    this, it would clopper `BENCHFILE` with the default bench file.
  * `make bench`: Use -10 priority with renice.
  * Improved `make_autolink`. Ensures that title is chunk with empty
    string rather than NULL, as with other links.
  * Added `clang-check` target.
  * Travis: split `roundtrip_test` and `leakcheck` (OGINO Masanori).
  * Use clang-format, llvm style, for formatting. Reformatted all source files.
    Added `format` target to Makefile. Removed `astyle` target.
    Updated `.editorconfig`.

[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:
    `&ngE;` should be rendered as "≧̸" (U+02267 U+00338), but it was
    being rendered as "≧" (which is the same as `&gE;`).
  * 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).
  * Don't allow link labels with no non-whitespace characters
    (jgm/CommonMark#322).
  * Fixed multiple issues with numeric entities (#33, Nick Wellnhofer).
  * Support CR and CRLF line endings (Ben Trask).
  * Added test for different line endings to `api_test`.
  * Allow NULL value in string setters (Nick Wellnhofer).  (NULL
    produces a 0-length string value.)  Internally, URL and
    title are now stored as `cmark_chunk` rather than `char *`.
  * Fixed memory leak in `cmark_consolidate_text_nodes` (#32).
  * Fixed `is_autolink` in the CommonMark renderer (#50).  Previously *any*
    link with an absolute URL was treated as an autolink.
  * Cope with broken `snprintf` on Windows (Nick Wellnhofer).  On Windows,
    `snprintf` returns -1 if the output was truncated. Fall back to
    Windows-specific `_scprintf`.
  * Switched length parameter on `cmark_markdown_to_html`,
    `cmark_parser_feed`, and `cmark_parse_document` from `int`
    to `size_t` (#53, Nick Wellnhofer).
  * Use a custom type `bufsize_t` for all string sizes and indices.
    This allows to switch to 64-bit string buffers by changing a single
    typedef and a macro definition (Nick Wellnhofer).
  * Hardened the `strbuf` code, checking for integer overflows and
    adding range checks (Nick Wellnhofer).
  * Removed unused function `cmark_strbuf_attach` (Nick Wellnhofer).
  * Fixed all implicit 64-bit to 32-bit conversions that
    `-Wshorten-64-to-32` warns about (Nick Wellnhofer).
  * Added helper function `cmark_strbuf_safe_strlen` that converts
    from `size_t` to `bufsize_t` and throws an error in case of
    an overflow (Nick Wellnhofer).
  * Abort on `strbuf` out of memory errors (Nick Wellnhofer).
    Previously such errors were not being trapped.  This involves
    some internal changes to the `buffer` library that do not affect
    the API.
  * Factored out `S_find_first_nonspace` in `S_proces_line`.
    Added fields `offset`, `first_nonspace`, `indent`, and `blank`
    to `cmark_parser` struct.  This just removes some repetition.
  * Added Racket Racket (5.3+) wrapper (Eli Barzilay).
  * Removed `-pg` from Debug build flags (#47).
  * Added Ubsan build target, to check for undefined behavior.
  * Improved `make leakcheck`.  We now return an error status if anything
    in the loop fails.  We now check `--smart` and `--normalize` options.
  * Removed `wrapper3.py`, made `wrapper.py` work with python 2 and 3.
    Also improved the wrapper to work with Windows, and to use smart
    punctuation (as an example).
  * In `wrapper.rb`, added argument for options.
  * Revised luajit wrapper.
  * Added build status badges to README.md.
  * Added links to go, perl, ruby, R, and Haskell bindings to README.md.

[0.19.0]

  * Fixed `_` emphasis parsing to conform to spec (jgm/CommonMark#317).
  * Updated `spec.txt`.
  * Compile static library with `-DCMARK_STATIC_DEFINE` (Nick Wellnhofer).
  * Suppress warnings about Windows runtime library files (Nick Wellnhofer).
    Visual Studio Express editions do not include the redistributable files.
    Set `CMAKE_INSTALL_SYSTEM_RUNTIME_LIBS_NO_WARNINGS` to suppress warnings.
  * Added appyeyor: Windows continuous integration (`appveyor.yml`).
  * Use `os.path.join` in `test/cmark.py` for proper cross-platform paths.
  * Fixed `Makefile.nmake`.
  * Improved `make afl`:  added `test/afl_dictionary`, increased timeout
    for hangs.
  * Improved README with a description of the library's strengths.
  * Pass-through Unicode non-characters (Nick Wellnhofer).
    Despite their name, Unicode non-characters are valid code points. They
    should be passed through by a library like libcmark.
  * Check return status of `utf8proc_iterate` (#27).

[0.18.3]

  * Include patch level in soname (Nick Wellnhofer). Minor version is
    tied to spec version, so this allows breaking the ABI between spec
    releases.
  * Install compiler-provided system runtime libraries (Changjiang Yang).
  * Use `strbuf_printf` instead of `snprintf`. `snprintf` is not
    available on some platforms (Visual Studio 2013 and earlier).
  * Fixed memory access bug: "invalid read of size 1" on input `[link](<>)`.

[0.18.2]

  * Added commonmark renderer: `cmark_render_commonmark`. In addition
    to options, this takes a `width` parameter.  A value of 0 disables
    wrapping; a positive value wraps the document to the specified
    width.  Note that width is automatically set to 0 if the
    `CMARK_OPT_HARDBREAKS` option is set.
  * The `cmark` executable now allows `-t commonmark` for output as
    CommonMark.  A `--width` option has been added to specify wrapping
    width.
  * Added `roundtrip_test` Makefile target.  This runs all the spec
    through the commonmark renderer, and then through the commonmark
    parser, and compares normalized HTML to the test.  All tests pass
    with the current parser and renderer, giving us some confidence that
    the commonmark renderer is sufficiently robust.  Eventually this
    should be pythonized and put in the cmake test routine.
  * Removed an unnecessary check in `blocks.c`.  By the time we check
    for a list start, we've already checked for a horizontal rule, so
    we don't need to repeat that check here.  Thanks to Robin Stocker for
    pointing out a similar redundancy in commonmark.js.
  * Fixed bug in `cmark_strbuf_unescape` (`buffer.c`).  The old function
    gave incorrect results on input like `\\*`, since the next backslash
    would be treated as escaping the `*` instead of being escaped itself.
  * `scanners.re`:  added `_scan_scheme`, `scan_scheme`, used in the
    commonmark renderer.
  * Check for `CMAKE_C_COMPILER` (not `CC_COMPILER`) when setting C flags.
  * Update code examples in documentation, adding new parser option
    argument, and using `CMARK_OPT_DEFAULT` (Nick Wellnhofer).
  * Added options parameter to `cmark_markdown_to_html`.
  * Removed obsolete reference to `CMARK_NODE_LINK_LABEL`.
  * `make leakcheck` now checks all output formats.
  * `test/cmark.py`:  set default options for `markdown_to_html`.
  * Warn about buggy re2c versions (Nick Wellnhofer).

[0.18.1]

  * Build static version of library in default build (#11).
  * `cmark.h`:  Add missing argument to `cmark_parser_new` (#12).

[0.18]

  * Switch to 2-clause BSD license, with agreement of contributors.
  * Added Profile build type, `make prof` target.
  * Fixed autolink scanner to conform to the spec. Backslash escapes
    not allowed in autolinks.
  * Don't rely on strnlen being available (Nick Wellnhofer).
  * Updated scanners for new whitespace definition.
  * Added `CMARK_OPT_SMART` and `--smart` option, `smart.c`, `smart.h`.
  * Added test for `--smart` option.
  * Fixed segfault with --normalize (closes #7).
  * Moved normalization step from XML renderer to `cmark_parser_finish`.
  * Added options parameter to `cmark_parse_document`, `cmark_parse_file`.
  * Fixed man renderer's escaping for unicode characters.
  * Don't require python3 to make `cmark.3` man page.
  * Use ASCII escapes for punctuation characters for portability.
  * Made `options` an int rather than a long, for consistency.
  * Packed `cmark_node` struct to fit into 128 bytes.
    This gives a small performance boost and lowers memory usage.
  * Repacked `delimiter` struct to avoid hole.
  * Fixed use-after-free bug, which arose when a paragraph containing
    only reference links and blank space was finalized (#9).
    Avoid using `parser->current` in the loop that creates new
    blocks, since `finalize` in `add_child` may have removed
    the current parser (if it contains only reference definitions).
    This isn't a great solution; in the long run we need to rewrite
    to make the logic clearer and to make it harder to make
    mistakes like this one.
  * Added 'Asan' build type. `make asan` will link against ASan; the
    resulting executable will do checks for memory access issues.
    Thanks @JordanMilne for the suggestion.
  * Add Makefile target to fuzz with AFL (Nick Wellnhofer)
    The variable `$AFL_PATH` must point to the directory containing the AFL
    binaries. It can be set as an environment variable or passed to make on
    the command line.

[0.17]

  * Stripped out all JavaScript related code and documentation, moving
    it to a separate repository (<https://github.com/jgm/commonmark.js>).
  * Improved Makefile targets, so that `cmake` is run again only when
    necessary (Nick Wellnhofer).
  * Added `INSTALL_PREFIX` to the Makefile, allowing installation to a
    location other than `/usr/local` without invoking `cmake`
    manually (Nick Wellnhofer).
  * `make test` now guarantees that the project will
    be rebuilt before tests are run (Nick Wellnhofer).
  * Prohibited overriding of some Makefile variables (Nick Wellnhofer).
  * Provide version number and string, both as macros
    (`CMARK_VERSION`, `CMARK_VERSION_STRING`) and as symbols
    (`cmark_version`, `cmark_version_string`) (Nick Wellnhofer).  All of
    these come from `cmark_version.h`, which is constructed from a
    template `cmark_version.h.in` and data in `CMakeLists.txt`.
  * Avoid calling `free` on null pointer.
  * Added an accessor for an iterator's root node (`cmark_iter_get_root`).
  * Added user data field for nodes (Nick Wellnhofer).  This is
    intended mainly for use in bindings for dynamic languages, where
    it could store a pointer to a target language object (#287).  But
    it can be used for anything.
  * Man renderer:  properly escape multiline strings.
  * Added assertion to raise error if finalize is called on a closed block.
  * Implemented the new spec rule for emphasis and strong emphasis with `_`.
  * Moved the check for fence-close with the other checks for end-of-block.
  * Fixed a bug with loose list detection with items containings
    fenced code blocks (#285).
  * Removed recursive algorithm in `ends_with_blank_line` (#286).
  * Minor code reformatting: renamed parameters.

[0.16]

  * Added xml renderer (XML representation of the CommonMark AST,
    which is described in `CommonMark.dtd`).
  * Reduced size of gperf entity table (Nick Wellnhofer).
  * Reworked iterators to allow deletion of nodes during iteration
    (Nick Wellnhofer).
  * Optimized `S_is_leaf`.
  * Added `cmark_iter_reset` to iterator API.
  * Added `cmark_consolidate_text_nodes` to API to combine adjacent
    text nodes.
  * Added `CMARK_OPT_NORMALIZE` to options (this combines adjacent
    text nodes).
  * Added `--normalize` option to command-line program.
  * Improved regex for HTML comments in inline parsing.
  * Python is no longer required for a basic build from the
    repository.