Releases: thephpleague/commonmark
Releases · thephpleague/commonmark
2.6.1
2.6.0
This is a security release to address potential denial of service attacks when parsing specially crafted,
malicious input from untrusted sources (like user input). See GHSA-c2pc-g5qf-rfrf for more details.
Added
- Added
max_delimiters_per_lineconfig option to prevent denial of service attacks when parsing malicious input - Added
table/max_autocompleted_cellsconfig option to prevent denial of service attacks when parsing large tables - The
AttributesExtensionnow supports attributes without values (#985, #986) - The
AutolinkExtensionexposes two new configuration options to override the default behavior (#969, #987):autolink/allowed_protocols- an array of protocols to allow autolinking forautolink/default_protocol- the default protocol to use when none is specified
- Added
RegexHelper::isWhitespace()method to check if a given character is an ASCII whitespace character - Added
CacheableDelimiterProcessorInterfaceto ensure linear complexity for dynamic delimiter processing - Added
Bracketdelimiter type to optimize bracket parsing
Changed
[and]are no longer added asDelimiterobjects on the stack; a newBrackettype with its own stack is used insteadUrlAutolinkParserno longer parses URLs with more than 127 subdomains- Expanded reference links can no longer exceed 100kb, or the size of the input document (whichever is greater)
- Delimiters should always provide a non-null value via
DelimiterInterface::getIndex()- We'll attempt to infer the index based on surrounding delimiters where possible
- The
DelimiterStacknow accepts integer positions for any$stackBottomargument - Several small performance optimizations
2.5.3
Changed
- Made compatible with CommonMark spec 0.31.1, including:
- Remove
source, addsearchto list of recognized block tags
- Remove
Full Changelog: 2.5.2...2.5.3
2.5.2
Changed
- Boolean attributes now require an explicit
truevalue (#1040)
Fixed
- Fixed regression where text could be misinterpreted as an attribute (#1040)
Full Changelog: 2.5.1...2.5.2
2.5.1
Fixed
- Fixed attribute parsing incorrectly parsing mustache-like syntax (#1035)
- Fixed incorrect
Tablestart line numbers (#1037)
New Contributors
- @jasonvarga made their first contribution in #1035
Full Changelog: 2.5.0...2.5.1
2.5.0
Added
- The
AttributesExtensionnow supports attributes without values (#985, #986) - The
AutolinkExtensionexposes two new configuration options to override the default behavior (#969, #987):autolink/allowed_protocols- an array of protocols to allow autolinking forautolink/default_protocol- the default protocol to use when none is specified
Changed
- Made compatible with CommonMark spec 0.31.0, including:
- Allow closing fence to be followed by tabs
- Remove restrictive limitation on inline comments
- Unicode symbols now treated like punctuation (for purposes of flankingness)
- Trailing tabs on the last line of indented code blocks will be excluded
- Improved HTML comment matching
Paragraphs only containing link reference definitions will be kept in the AST until theDocumentis finalized- (These were previously removed immediately after parsing the
Paragraph)
- (These were previously removed immediately after parsing the
Fixed
- Fixed list tightness not being determined properly in some edge cases
- Fixed incorrect ending line numbers for several block types in various scenarios
- Fixed lowercase inline HTML declarations not being accepted
New Contributors
- @svenluijten made their first contribution in #986
Full Changelog: 2.4.4...2.5.0
2.4.4
Fixed
- Fixed SmartPunct extension changing already-formatted quotation marks (#1030)
Full Changelog: 2.4.3...2.4.4
2.4.3
Fixed
- Fixed the Attributes extension not supporting CSS level 3 selectors (#1013)
- Fixed
UrlAutolinkParserincorrectly parsing text containingwwwanywhere before an autolink (#1025)
New Contributors
- @nfreader made their first contribution in #1016
- @macbookandrew made their first contribution in #1025
- @xavierlacot made their first contribution in #1013
Full Changelog: 2.4.2...2.4.3
2.4.2
Fixed
- Fixed declaration parser being too strict
FencedCodeRenderer: don't addlanguage-to class if already prefixed
New Contributors
- @sergiy-petrov made their first contribution in #997
- @clustermin made their first contribution in #988
- @peter279k made their first contribution in #996
Full Changelog: 2.4.1...2.4.2