Age | Commit message (Collapse) | Author | Files | Lines |
|
Keep the old ChangeLog files as ChangeLog.old, and distribute them.
Automatically genereate the ChangeLog from “git log”. And update the
information for translators.
|
|
|
|
This fixes “make distcheck”, as t.tmp is a read/write directory it has
to be used under builddir, which in most cases happens to be the same as
srcdir.
|
|
Otherwise gettext would return the special msgid for the catalog.
|
|
The C locale is supposed to only get ASCII strings, for non-ASCII
strings there are things like the en_US or en_GB locales.
|
|
The build was failing when xgettext was trying to scan the non-existent
file scripts/cleanup-info.pl.
|
|
|
|
To better support packages of libraries that have different names
between architectures, offer the possibility to not hardcode the package
name in the symbols file by using #PACKAGE#. This marker is then
substituted by dpkg-gensymbols when the symbols files are installed inside
the binary package.
|
|
Add one more sanity check verifying that <link> and <path> are different
(cf #509667 for a sample).
|
|
Make sure that all messages that are likely to appear in the
output as part of --install and --remove call are identified
as coming from update-alternatives. Factorize the logic of output and
verbosity in some standard function (info, warning, verbose).
|
|
The ustar format has been supported up to now except for the long names
which use the prefix field that needs to be prepended to the name field
if non empty.
Closes: #474092
|
|
Recognize old tar, GNU tar and ustar formats. Abort on ustar with non
empty Prefix field, as we don't properly handle the long names yet.
Failure for PAX archive is already being handled when acting on the
typeflag.
|
|
Fix the function to use PKGPFIELD to assign to the correct struct
member instead of directly assigning to the essential member. This also
makes this function match its write counterpart w_booleandefno's generic
behaviour.
|
|
The field should have been already initialized to 0 by
blankpackageperfile, so no need to assign again.
|
|
|
|
Add new option -I, --iosched <class[:priority]>.
Closes: #443535
Signed-off-by: Guillem Jover <guillem@debian.org>
|
|
Reading the files database can take a while on machines with slow disks
and an empty cache. To make the wait more tolerable try to display a
progress indicator if the output is a terminal.
Based-on-patch-by: Romain Francoise <rfrancoise@debian.org>
|
|
Initial support to report progress from dpkg. For now it only changes
behaviour if printing on a tty, later on it could be made pluggable so
that external applications can monitor the internal progress on
expensive dpkg actions.
Based-on-patch-by: Romain Francoise <rfrancoise@debian.org>
|
|
This option overrides the default log file /var/log/dpkg.log.
It is needed to make the update-alternatives test-suite succeed
when run with fakeroot because without override the test-suite would
try to write to /var/log/dpkg.log and fail. Hence adjust
the test-suite accordingly.
|
|
--force
update-alternatives should not replace real files by symlinks. When
--install is called, the alternative link should either not exist
or already be a link. It that's not the case, it will simply skip
installing that link and display a warning. However if --force is given,
it will (try to) replace the file.
Ensure that behaviour within the test suite.
|
|
Since alternatives are repaired every time they are reconfigured,
the user can use --all to reconfigure them all with the current choice
and repair everything. Document this in the manual pages and ensure
that it really works that way by returning a valid choice.
Closes: #250258, #395556
|
|
The option --set-selections is the counterpart of --get-selections,
it reads configuration on standard input and reconfigures alternatives
accordingly.
|
|
The new option list all master alternatives and their status
in a format simple to parse. It's also a practical way to know
the name of all available alternatives. Closes: #273406, #392429
|
|
All slave links are optional in the sense that they are only installed
if the currently associated file does exist. This allows most
update-alternatives call to succeed even if the admin removed
documentation files or other optional files. Closes: #143701
A message is still displayed by default in this case to warn the
user that something uncommon has been detected.
Adjusted the test suite to verify this behaviour.
|
|
Check that links and alternative paths are absolute. Check that the
master alternative path exists. Forbid "/" and spaces in alternative names.
Closes: #423176
Add corresponding tests in the test-suite.
|
|
Fail if --install tries to reuse links owned by other alternatives, or
if the alternative type (slave/master) doesn't match the information
already available. Closes: #342566
Modify test-suite to verify this behaviour.
|
|
Add logging statements in update-alternatives for all calls that can
modify the status of the link group. Also log explicitely alternative
removal, auto-reparation and update. Closes: #445270
|
|
The vocabulary used was misleading and not really uniform. Try to
standardize and make it match with the vocabulary used in the source
code too.
Includes minor fixes to make it match the behaviour of the current
implementation.
|
|
Full rewrite of update-alternatives. It was badly needed so that
we can actually understand its behaviour and implement new features
on top of it. This commit doesn't add any new feature but still improves
the behaviour of --config: it has a nicer output (with choices sorted
alphabetically, closes: #437060) and it accepts a path as well as the
index of the choice (to make it easier to write non regression tests for
this action). The test-suite is adjusted accordingly.
|
|
Reset umask to 0022 to ensure that files created in the DEBIAN directory
have sane permissions. Closes: #516481
|
|
|
|
* scripts/Makefile.am: Drop cleanup-info.pl script.
* man/Makefile.am, man/po/po4a.cfg: Drop cleanup-info(8) manual page.
* README.feature-removal-schedule: Update information accordingly.
|
|
* scripts/Makefile.am: Move update-alternatives, dpkg-divert and
dpkg-statoverride to /usr/bin as they can be useful to users since
they have --display or --list options that do not require
administrative rights.
* Makefile.am (EXTRA_DIST): Add debian/dpkg.links
|
|
* scripts/dpkg-buildpackage.pl: Add new option --target/-T to
allow a specific debian/rules target to be run. The new --as-root
option forces the target to be run with root rights.
Add support for the two common syntaxes for long
options ("--admindir=…" and "--admindir …").
* man/dpkg-buildpackage.1: Update documentation accordingly.
|
|
|
|
|
|
|
|
Remove the missplaced version constructor from inside
assert_version_support, and instead just pass an already initialized
structure. This also allows making the version structures non-static.
|
|
|
|
* scripts/t/900_update_alternatives.t: New test-suite for
update-alternatives.
* scripts/Makefile.am: Add the previous file.
|
|
* scripts/update-alternatives.pl: Deal with empty
files in the administrative directory by ignoring them.
|
|
* scripts/update-alternatives.pl: Ensure that a non-existant
alternative result in automatic mode configuration.
|
|
* scripts/update-alternatives.pl: Change set_links() to
also reset the main alternative link. Add a parameter to
make it quiet.
When --install reinstalls the current and manually selected entry,
let it recreate the symlinks in case something has gone wrong.
|
|
* scripts/update-alternatives.pl: Remove the link group when
the last choice is removed, even if we are in manual mode.
* man/update-alternatives.8: Update the documentation
accordingly to not be as strong about not touching links in manual
mode.
|
|
* scripts/update-alternatives.pl (set_links): Remove slavelinks
if needed as well. We create slave links automatically, we should
remove them in the same way. Otherwise --config or --set will let
dangling symlinks on the system.
|
|
|
|
Ar member names, types of tar archives and data.tar members.
|
|
Closes: #509312
|
|
|
|
* scripts/update-alternatives.pl (rename_mv): Integrate logic to
not fail if ENOENT is the reason why the rename failed.
The underlying problem in the previous implementation is not
immediately obvious but rename() can fail for multiple reasons
at the same time. In #99870 we have a case where rename() fails
with EXDEV and the system("mv",…) call fails due to ENOENT. In that
case, rename_mv() fails but the check $! == ENOENT also fails
when in fact it should not because ENOENT is really the reason
why rename_mv() failed.
Verifying the existence of the source file first is a sanity check
that avoids running mv when we know that it's going to fail and clutter
the screen with an undesired error message.
|