Changelog
dgit (1.1) unstable; urgency=medium
Bugfixes:
* When source package contains things called .git (even files, and even
in subdirectories), remove them. Closes:#793671.
* Work around curl -sS -I printing `HTTP/1.0 200 Connection established'
before the actual header, so dgit works with https_proxy set (!)
* --new is needed for read access to packages in NEW, too. Document
this, and make it work properly.
* Work around #793471 (madness with $SIG{__WARN__} and Perl's system
builtin): move $SIG{} setting into setup_sigwarn in Dgit.pm, and
check getppid.
* When invoking git-buildpackage via dgit gbp-build, consider our
command line arguments when massaging the dpkg-buildpackage arguments,
so that we don't end up giving dpkg-buildpackage contradictory
instructions.
* Cope with new git-buildpackage which provides gbp, rather than the
eponymous command, on PATH.
Configurability:
* Honour dgit-distros.DISTRO.cmd-CMD and .opts-CMD. Closes:#793427.
* Make configuration able to prevent dpkg-mergechangelogs setup.
* Provide dgit setup-new-tree (like dpkg-setup-mergechangelogs
but only does it if not disabled in config).
* Set up git user.email and user.name from distro access config
or DEBEMAIL/DEBFULLNAME. Closes:#793410.
* When key to use not specified any other way, use the debian/changelog
trailer line. Closes:#793423.
* Honour --git= (mostly).
Documentation:
* Fix some manpage typos. [ Richard Hartmann ]
* Manpage said that --clean=check was -wn but that is --clean=none;
correctly document that --clean=check is actually -wc.
* Document that up to -DDDD (not just -DD) is meaningfully different.
* Document that -cname=value applies only for this run.
* Improve manpage comment about defining a new distro.
* Document that --quilt=linear is the default for Debian.
* Fix a formatting problem in --build-products-dir= doc.
* In manpage, do not seem to imply that NMU should be of only one
new commit.
* Qualify to Debian the manpage comment about how to do NMU.
* In discussion on how to start using dgit when already using git, do
not imply/assume that existing git history will have identical trees
to dgit history.
* Remove stray sentence in config section of manpage.
* Manpage: Clarify wording of readonly config.
* Manpage: Better cross-references for -k and keyid.
* dgit(7): No longer say that dgit-repos lives on Alioth.
Improvements:
* Introduce more sophisticated protocol negotiation for rpush.
* Do not quote `:' in shellquote.
* Print a supplementary message when push fails, giving advice to
the user about how to retry. Closes:#793144.
* Slurp in entire git config, for better performance.
* Rename `git-build' operation to `gbp-build' to make it clearer what
it's for. Keep the old name as an alias.
* Show `dgit sbuild' in usage message.
* When we are using dpkg-buildpackage to clean before using it to also
do the build, let it do its cleaning thing as part of its run, rather
than running it twice. When we are _not_ supposed to be using
dpkg-buildpackage to clean, but we are running it to do the build,
pass -nc.
* Also suppress spurious runs of the clean target when building using
git-buildpackage.
* When exec fails, always print the program name in the error message.
Infrastructure:
* Infrastructure: Get mirroring right for fresh repos of existing
packages (!)
Packaging, cleanups, debugging and test suite:
* Fix Vcs-Git and Vcs-Browse to refer to chiark. (The dgit-repos on
alioth aren't suitable right now because the master there can
currently only be updated with an actual upload, ie dgit push.)
* Make warnings fatal in dpkg-repos-admin-debian, dgit-ssh-dispatch
(using setup_sigwarn).
* Make warnings fatal in tstunt/dpkg-parsechangelog (directly).
* Test suite: Honour DGIT_TEST_DEBUG=''.
* With -DDDD, print out all gitcfg references (copious!)
* Fix a debug message in the obsolete sshpsql archive access driver.
* Test suite: More automatic enumeration of tests.
* Test suite: Provide tests which check that all our various build
operations run the right targets as expected (ie, that we are massaging
the arguments to dpkg-buildpackage, and suppressing our clean target,
etc., correctly).
-- Ian Jackson <email address hidden> Mon, 27 Jul 2015 16:34:31 +0100