Changes¶

Pre-1.0 Series¶

Development

doc Prepared the first version of a package documentation generated with Sphinx (HTML, PDF (LuaLaTeX) and some manual pages).

breaking Combine Unix epoch and ISO timestamp values into one record. The parser recognizes the previous format also.
breaking Option --ignore-symlink-type is now the default. Deactivate with --no-ignore-symlink-type.
breaking Shorten some record markers in the “tabular” output style: ERROR: into ERR: and WARNING: into WARN:.

Because no parser existed previously the compatibility break is somewhat limited.
breaking In the “tabular” output style errors regarding size numbers are sequences of / characters instead of ? previously.

Because no parser existed previously the compatibility break is somewhat limited.
feature Implemented a parser for the “tabular” output style.
feature Allow to put warnings into the output. May be dectivated with option without-warnings.
feature Implemented option --with-notes to optionally put detailed data about ignored files or directories into the output.
bugfix In Python 2: the “surrogateescape” error handler (introduced in v0.11) is not called when is should be called: some illegal UTF-8 is accepted while it should be treated as illegal. This is fixed now.
bugfix Obey --ignore-symlink-type also for symbolic links on the command lines.

feature Implemented option --ignore-symlink-type: when not following symlinks to directories and files do also not try to determine the link target’s type (which can be done only when resolving the link).

breaking Message digests for directories that contain non-ASCII filenames changed because of a new filename normalization algorithm.

The digest values are now independent of the --utf8 or --fsencoding settings. These settings now just determine how filenames in the output are encoded.

These are tested on a POSIX system using Python versions 2.7, 3.4, 3.5, 3.6, 3.7, 3.9, 3.11 and 3.14 using different locale settings.
breaking On most handled errors do not use the concrete OS errno for digesting but an OS independent generic one. Note that this may change message digests in the output.
breaking Some record formats of a “treesum” file changed to be more consistent.
breaking Make --utf8 the default instead of --fsencoding.
breaking Computation of the CRC is now independent of the native OS lineendings that are in effect. Also lines with no records (whitespace only lines, comment lines using ; or #) do not contribute to the CRC.
feature Implemented --rfc1924 and --base85 for shorter digest representations in the output. Also implemented --base32 for completeness.
bugfix In some cases the TAINTED property of a message digest was not propagated to parent directories properly.

misc Modernize the package builds: switched to a PEP 517 and PEP 518 compatible build system (using pyproject.toml).

For older Python versions there is a branch build-2.7 in Mercurial that allows builds with older Python versions (using the “legacy” setup.py).

bugfix Fix an oversight that led to crashes on Python version before 2.7.9 (where hashlib.algorithms_available and hashlib.algorithms_guaranteed are not available).
bugfix Optimize CR/LF handling on Windows when using I/O redirection.
misc Make ERROR outputs more consistent.

This is the first non-private release.

Combine Unix epoch and ISO timestamp values into one record line.
--ignore-symlink-type is now the default.
Shorten some record markers in the “tabular” output style: ERROR: into ERR: and WARNING: into WARN:.
In the “tabular” output style errors regarding size numbers are sequences of / characters instead of ? previously.

Message digests for directories that contain non-ASCII filenames changed because of a new filename normalization algorithm.
Changed the way how OSError exceptions contribute to message digests.
Some record formats of a “treesum” file changed to be more consistent.
Computation of the CRC is now independent of the native OS lineendings that are in effect. Also lines with no records (whitespace only lines, comment lines using ; or #) do not contribute to the CRC.