2020-05-16: C TAP Harness 4.7
This is a small bug fix release to my testing framework for C packages.
It picks up a change to the test suite so that it won't break when
C_TAP_VERBOSE is already set in the environment, and fixes new
compilation warnings with GCC 10.
You can get the latest release from the
C TAP Harness distribution page.
2020-05-16: rra-c-util 8.2
This release of my general utility libraries and support code includes a
large grab bag of fixes and improvements.
portable/system.h now defines explicit_bzero in terms of
memset if it is not available. The memset version is
unlikely to have the same security properties since the compiler may
optimize it away, but that allows me to use explicit_bzero to erase
security data where it is available.
For packages with Kerberos tests, generating a test krb5.conf file
now works properly even if the system krb5.conf file does not set a
default realm, and a krb5.conf file dropped into the test
configuration directory now works properly. Thanks to Jeffrey Hutzelman
for the latter fix.
For packages with PAM modules, the ENTRY and EXIT logging
macros can now be used like function calls, and portable/pam.h now
defines PAM_MAX_RESP_SIZE if it isn't defined.
Header ordering in some of the portability socket code has been restored
to compatibility with a few ancient UNIX systems. This was accidentally
broken by the clang-format reformatting. Thanks to Julien ÉLIE for the
fix.
A few bugs in the test for SPDX license identifiers have been fixed.
Finally, this release fixes warnings with Clang 10 and GCC 10.
You can get the latest release from the rra-c-util distribution page.
2020-05-16: DocKnot 3.04
This is a relatively small feature release of my tool for managing
software documentation and releases.
I'm slowly moving all of my packages from Travis-CI to GitHub Workflows
for automated CI. GitHub Workflows is much easier to configure and
control, and I've been a bit worried about the future of Travis-CI since
their acquisition. It seems unlikely that GitHub Workflows is going
anywhere.
It would be nice to use a fully free software solution for CI, but there
doesn't seem to be anything out there that's nearly as easy and
straightforward to use, and I have neither the time nor the energy to
cobble something together myself. The configuration is fairly
straightforward and should be portable to any fully free solution that
might materialize in the future.
Anyway, as part of that migration I needed to make some changes to DocKnot
to generate status badges from GitHub Workflows instead of Travis-CI.
This release includes those changes. There is a backward-incompatible
change to make the semantics of the package metadata a bit more
straightforward: vcs.travis needs to be changed to
vcs.status.travis.
You can get the latest release from the DocKnot distribution page. Debian packages have been uploaded to my
personal repository. I plan on
uploading DocKnot to Debian proper once I change the metadata format to
use YAML instead of relaxed JSON.
2020-05-17: krb5-strength 3.2
krb5-strength provides password strength checking for Kerberos KDCs
(either MIT or Heimdal), and also provides a password history
implementation for Heimdal.
This release adds a check-only mode to the heimdal-history command
to interrogate history without modifying it and increases the default hash
iterations used when storing old passwords. explicit_bzero is now
used, where available, to clear the memory used for passwords after
processing. krb5-strength can now optionally be built without CrackLib
support at all, if you only want to use the word list, edit distance, or
length and character class rules.
It's been a few years since the previous release, so this release also
updates all the portability code, overhauls valgrind testing, and now
passes tests when built with system CrackLib (by skipping tests for
passwords that are rejected by the stronger rules of the embedded CrackLib
fork).
You can get the latest release from the
krb5-strength distribution page.
New packages will be uploaded to Debian unstable shortly (as soon as a
Perl transition completes enough to make the package buildable in
unstable).
All posts