0.60.8.1

Changes 0.60.8 to 0.60.8.1

• Fix memory leak in suggestion code introduced in 0.60.8.
• Various documentation fixes.
• Fix various warnings when compiling with -Wall.
• Fix two buffer overflows found by Google’s OSS-Fuzz.
• Other minor updates.

0.60.8

Please note that this version has the potential to break some applications that use null-terminated UCS-2 and UCS-4 encoded strings in order to fix a potentially unbounded buffer over-read. Very few applications do this however. For additional information please see "Upgrading from Aspell 0.60.7" in the manual.

Major changes from 0.60.7 to 0.68.8:

• Prevent a potentially unbounded buffer over-read by no longer supporting null-terminated UCS-2 and UCS-4 encoded strings with the original C API.
• Ensure that possible typos are listed before other suggestions when typo analysis is used. Also fix a bug so that suggestions that split a word using a space or hyphen are not always first.
• Add Markdown filter.
• Add new 'wordlists' option, which is a list of UTF-8 files that contain additional words to accept.
• Add new 'camel-case' option, which enables support for checking camelCase words.
• Sort personal and replacement dictionaries.
• Change 'ultra' suggestion mode to only find words that are within one-edit distance or have the same soundslike.

Full Changelog

0.60.8-20190921

Changes from last snapshot (20190817):

• Add new 'wordlists' option, which is a list of UTF-8 files that contain additional words to accept.
• Provide support for checking camelCase words.
• The 'clean' option and command will no longer split a word.

Other major changes from Aspell 0.60.7:

• Add Markdown filter.
• When typo analysis is used, ensure that possible typos are listed before other suggestions. Also fix a bug so that suggestions that split a word using a space or hyphen are not always first.
• To prevent a potentially unbounded buffer over-read Aspell no longer supports null-terminated UCS-2 and UCS-4 encoded strings with the original C API.

0.60.8-20190817

Due to the discovery of a potential buffer over-read in the API the next version of Aspell (0.60.8) and this snapshot may break applications that pass null-terminated UCS-2 or UCS-4 strings to the C API. More

Changes so far (from Aspell 0.60.7):

• Add Markdown filter.
• When typo analysis is used, ensure that possible typos are listed before other suggestions. Also fix a bug so that suggestions that split a word using a space or hyphen are not always first.
• Change ‘ultra’ suggestion mode to only find words that are within one edit distance or have the same soundslike.
• To prevent a potentially unbounded buffer over-read Aspell no longer supports null-terminated UCS-2 and UCS-4 encoded strings with the original C API. See Upgrading from Aspell 0.60.7.
• Fix a bug in ‘AspellDocumentChecker’ that prevented it from working with UCS-2 and UCS-4 encoded strings.
• Implement ‘aspell filter’ command.
• Remove unused ‘sug-edit-dist’ option.
• In order to work with the new Markdown filter ‘AspellDocumentChecker’ now expects the document a line at a time. If the document is split on white space characters instead, nothing will break, but new filters such as the Markdown filter may give incorrect results.
• Fix various crashes and other problems found by Google’s OSS-Fuzz.

0.60.7

This is primary a maintenance release. Major changes from 0.60.6.1 include:

• Add partial support for recognizing the Unicode apostrophe (’) in words. In particular Aspell will accept the Unicode apostrophe when the language uses an ISO Latin charset that doesn’t already have a Unicode apostrophe. For now, Aspell will still use the ASCII version in suggestions.

• Fix a bug that caused Aspell to crash with a SEGFAULT when built with mingw-w64.

• Fix a bug which caused Aspell to crash when passing in a null string to almost any of the C API functions. This should not happen if the size is also zero as the pointer should never be derefrenced.

• Detect when a dictionary compiled on a 32-bit machine is used on a 64-bit one (and vise versa), as due to an oversight, compiled dictionaries depend on more than the endianness. Also added a compile time option to remove this dependency, but at the cost of breaking compatibility with already compiled dictionaries on 64-bit systems.

Full Changelog