| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The history of tests is a little confusing. At one point, Niels updated
dh_installsystemd to install units to usr d70caa69c64b ("Make systemd
tools use usr/lib"), but he reverted that later 3083286de05f ("Revert
change to install systemd services in usr/lib"). That revert was
incomplete leaving the confusing "lib -> usr/lib" comments.
The "with both no-empty" case is now redundant after the reversion and
I'm therefore deleting it. As for the others, replace the comment with a
more useful one.
Signed-off-by: Helmut Grohne <helmut@subdivi.de>
|
|
|
|
|
|
|
|
|
|
|
| |
Since very long, systemd supports units in both /lib/systemd and
/usr/lib/systemd. As the /usr-merge progresses, we need debhelper to
also understand units in /usr/lib/systemd and this is what this commit
does. It does not cause debhelper to actually install files to /usr
rather than /lib and merely recognizes them.
Closes: #1041159
Signed-off-by: Helmut Grohne <helmut@subdivi.de>
|
|
|
|
|
|
|
|
|
| |
The problem that prompted this change was not caused by debhelper but
by fakeroot (#1024544) and the change in debhelper would not even
function as a work around in this case. Accordingly, the extra
complexity is not worth it and has been reverted.
Gbp-Dch: Full
|
|
|
|
|
|
|
|
|
|
| |
Previously, they would skip the `chown` call and rely on `dh_fixperms`
or `dpkg-deb` to fix the ownership. However, that could cause issues
when the helper was run after `dh_fixperms` and install files that
`dpkg-deb` would not reset ownership of.
Gbp-Dch: full
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
|
|
| |
This is an alternative to `install_dir` that never does chown (and is
not printed with `--verbose`). It is intended for creating
directories that are not part of a binary package.
Note that at the moment `install_dir` does not do chown either,
which is a bug (causing #1024261).
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
|
| |
Currently only one entry is preserved. This makes seldomly updated packages
(for example packages refreshed only every second Debian release) end up
with an excessively trimmed 1-entry changelog.
To prevent this issue, always keep at least 4 entries.
Closes: #1021467
|
|
|
|
|
|
|
|
|
|
|
| |
With the recent changes to dh_installinit this is no longer necessary as
systemd-sysv-generator will automatically ignore any SysV init scripts
that aren't marked as executable.
As masking system units on package removal can have undesired side
effects, remove this code.
Closes: #1021027
|
|
|
|
|
|
|
|
|
|
|
|
| |
The `dh_installchangelogs` tests generate fake timestamps using
the locale-abiding function `Time::Piece->strftime`.
Timestamps generated in a non-English environment will thus not be
dch-compliant and the tests will fail.
To fix this, use `setlocale` to run all `dh_installchangelogs` tests
with the `C.UTF-8` locale.
Fixes: #1021088
|
|
|
|
| |
Gbp-Dch: Ignore
|
|
|
|
| |
In case `notrimdch` is set, behave as if `--no-trim` had been passed.
|
|
|
|
|
|
|
|
|
|
| |
Changelogs of older and heavily developed packages can reach sizes of
hundreds of kilobytes. Even after compression, such big changelogs are
a waste of space and bandwidth worldwide.
This commit makes `dh_installchangelog` trim changelogs to include only
entries more recent than the release date of _oldstable_ (currently
Debian 10 Buster), unless the `--no-trim` option is passed.
|
|
|
|
|
|
|
|
|
|
|
| |
If a filename has multiple dot-digit sequences, the regex fails to
extract the last one, which is the section.
For example the section of libmanpage.so.9.2.3 is extracted as 9.2.3.
Fix the regex to match on the last dot and single digit
in the file name and letters following it.
This ensures we don't glob up 1.2.3, and also support .3perl and such.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When a manpage refers to a shared library, dh_installman detects the .so
as the language and moves the manpage into the wrong directory.
Eg:
/usr/share/man/man8/libnss_myhostname.so.2.8
gets processed as
/usr/share/man/so/man8/libnss_myhostname.so.2.8
Signed-off-by: Luca Boccassi <bluca@debian.org>
|
| |
|
|
|
|
| |
This reverts commit 5b27a5414e85d8ae6c6de8c8dc97cd0fa1897f53.
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
| |
init-system-helpers from v1.61 and systemd from v250 support
iterating over all user instances, so it's now possible to start/stop
user units on install/upgrade/remove.
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
By default autoconf will check for unknown --disable/enable/with/withouth-*
options and will emit a warning on standard error. This can be tuned by
the upstream maintainer with the AC_DISABLE_OPTION_CHECKING macro, to
make these fatal, or to not perform the checks by default.
Given that we unconditionally pass possibly unknown options to the
configure scripts, we should disable these checks to avoid emitting
warnings, which are at least confusing, and at worse might trip over
CI and test suite that check for output on standard error as markers
for failure.
Signed-off-by: Guillem Jover <guillem@debian.org>
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
| |
Thanks: Guillem Jover <guillem@debian.org>
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
| |
There was a bug in the original code that made the us use the highest
compat rather than the "current" compat. This hid a bug in the
dh_systemd.t test case, which has been corrected by only letting it
run for compat 10.
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
| |
Extracted from the patch proposed in https://bugs.debian.org/934893.
Signed-off-by: Niels Thykier <niels@thykier.net>
Signed-off-by: Simon McVittie <smcv@debian.org>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
| |
Previously, we relied on a command being inserted in the -arch
sequence to be ordered correctly. With this rewrite, a command added
only to the -indep sequence will still appear in the right order.
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commit [e5fc959e], resolving #462389 changed behaviour of `--name' option of
dh_installinit. Before this change, if `debian/{package}.{name}.init' is missing,
this option was silently ignored. This change made it error.
This change was incorrect, since it demanded presence of `debian/{package}.{name}.init'
file for /every/ binary package.
This commit instead throws error only if `debian/{package}.{name}.init' does
not exist for /all/ binary {package} names. Regression test is added.
Closes: #932073
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, `dh_installinit' silently did nothing, when --name option
was passed, but initscript debian/<package>.<name>.init was not found.
In almost all cases, explicit --name means that package maintainer meant
to install init script. If it is not present, it is bug, and must not be
hidden. Now, error is reported in this case. (Closes: #462389)
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
|
|
|
|
| |
Add a new 'dh_installsystemduser' helper responsible for istalling
package maintainer supplied systemd user instance units and to produce
postinst and postrm maintiner scripts code blocks to appropriately
enable, mask and disable units when the package is installed,
upgraded, or removed.
|
|
|
|
|
| |
The former calls /bin/pwd, while the latter uses the getcwd syscall
directly. This eliminates some forks and execs from every build.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Have dh_installinit use the new --skip-systemd-native parameter and
let dh_installsystemd always generate autoscript snippets for systemd
services.
This ensures that dh_installsystemd's snippet will be used for
starting the services and will among other ensure that services
are properly unmasked before started.
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|
|
|
|
| |
Signed-off-by: Niels Thykier <niels@thykier.net>
|