Merge branch 'release' into 'main'

Release 1.3.3

See merge request iwj/hippotat!75

Release 1.3.3

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'clippy' into 'main'

Fix clippy lints and add clippy to CI

See merge request iwj/hippotat!74

Merge branch 'no-verbose-ci' into 'main'

CI: Remove --verbose from cargo invocations

See merge request iwj/hippotat!73

Add clippy job using today's stable

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Suppress a lint for let _: Future = ...

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Suppress a lint for .get().is_some()

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

clippy: consistently allow complexity, rather than individual lints

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

CI: Remove --verbose from cargo invocations

This has IME never been useful.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'msrv' into 'main'

Bump MSRV to 1.85

See merge request iwj/hippotat!71

Bump MSRV to 1.85

This package doesn't build with dependencies from bookworm (currently
oldstable) any more.

And, in our MSRV test, test the actual MSRV compiler.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'ci-image' into 'main'

CI: Update most jobs to use trixie

See merge request iwj/hippotat!72

CI: Update most of the bookworm-using jobs

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

CI: Remove duplicate image: definition

Apparently gitlab doesn't spot this?

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'no-macro' into 'main'

Replace proc macro with normal code, and make hippotat-macros into a tombstone

See merge request iwj/hippotat!70

No longer depend on hippotat-macros

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Make hippotat-macros into a tombstone

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Remove a now-redundant ampersand

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Format lines_crlf call sites properly for new approach

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'cleanup-race' into 'main'

tests: Add a sleep to try to make tests more reliable

See merge request iwj/hippotat!68

LitLinesCrlf rework

Use the indentation of the first nonempty line, rather than trimming
all whitespace from the front of every line.  That will let us do
indented lines if we want.

Cons rather less.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Adjust LitLinesCrlf input string literals for consistent indentation

This will let us avoid trim, and instead trim a controlled amount.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

LitLinesCrlf: expect input without the final empty line

Before we got rid of the macro, this was handled by the way the `{}`
were embedded in the format string.

Then it was present in the source, stripped by LitLinesCrlf and then
added back by lines_crlf.  Now it is only handled by lines_crlf and
everything is reasonably straightforward.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Move code for into_crlfs macro to client.rs

This abolishes the macro.  It will allow us to abolish our macro crate
entirely.

Right now the version in client.rs is a bit odd and flabby.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: Add a sleep to try to make tests more reliable

Hopefully it will avoid this kind of thing (seen in ci.debian.net):

  101s + fuser -Mkm /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay
  101s /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay:  2659rce  2661rce  2699rce  2742rce  2772rce  2775rce  2777rce  2778rce  2779rce  2780rce  2781rce  2782rce
  101s + umount /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay/proc
  101s umount: /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay/proc: target is busy.
  101s + :
  101s + umount /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay/dev
  101s umount: /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay/dev: target is busy.
  101s + :
  101s + umount /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay/home
  101s /tmp/autopkgtest.lrleaj/build.Yox/src/adt/../test/tcommon: line 51:  2741 Killed                     $exec ip netns exec hippotat-t-$tname-$client_server "$@"
  101s + umount /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay
  101s umount: /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay: target is busy.
  101s + :
  101s + umount /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-pidns
  101s + test -d /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay
  101s + rmdir /tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay
  101s rmdir: failed to remove '/tmp/autopkgtest.lrleaj/autopkgtest_tmp/server-overlay': Device or resource busy
  101s /tmp/autopkgtest.lrleaj/build.Yox/src/adt/adt-full: line 1:  2529 Killed                     unshare --fork --pid="$tmp"/$cs-pidns chroot "$tmp"/$cs-overlay sh -ec '
  ...

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'update' into 'main'

Routine depedency updates

See merge request iwj/hippotat!67

Merge branch 'minimal-nightly' into 'main'

CI: Update pinned nightly for minimal-versions-nightly

See merge request iwj/hippotat!69

Allow use of sha2 0.11.x, and update to that

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update (minimal versions)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

CI: Update pinned nightly for minimal-versions-nightly

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'update' into 'main'

Dependency updates.  Allow reqwest 0.13.x.  Release 1.3.2.

See merge request iwj/hippotat!66

Release 1.3.2

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Makefile: Update build instructiosn for some places the version appears

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Allow use of reqwest 0.13.x (lockfile)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Allow use of reqwest 0.13.x

Closes: #1131578
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update (minimal versions)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'packaging' into 'main'

Misc tidying

See merge request iwj/hippotat!65

Document 1.3.1 and finalise changelog

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Add debian/source/options to declare native source format

For use by git-debpush.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Bump crate version (including macros) to 1.3.1

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'changelog' into 'main'

changelog: Document and finalise 1.3.0

See merge request iwj/hippotat!64

Bump crate versions to 1.3.0

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

changelog: Document and finalise 1.3.0

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'mr-57-combine' into 'main'

Add a systemd unit file for hippotatd.service (combined MR)

See merge request iwj/hippotat!63

HACKING.md: Explain what the weird autopkgtest rune is for

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: Run systemd debugging commands on failure

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: Tell systemd that we're running hippotatd as root

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'main' into mr-57-combine

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'autopkgtest' into 'main'

autopkgtests: Refactoring, preparatory to testing both systemd and sysvinit

See merge request iwj/hippotat!62

HACKING.md: Instructions for qemu-based autopkgtest

Mostly for my future self.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtests: Break out test-service-startup-main

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtests: Break out test-service-startup-prep

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtests: Fix a typo

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'bjh21-typos' into 'main'

Fix typos in documentation

See merge request iwj/hippotat!60

Fix typos in config.md

Assisted by M-x ispell-buffer.

Fix typos in install.md

Assisted by M-x ispell-buffer.

Add a systemd unit file for hippotatd.service

This mostly replicates the functionality of the existing init script.
It uses the same /etc/default/hippotatd as the init script, but loads it
using EnvironmentFile=.  That means that variable assignments in that
file will work, but not more complicated shell constructs.

systemd doesn't have a convenient way to set the service user from an
environment variable, so users who have set USER in
/etc/default/hippotatd will instead need to create a drop-in file under
/etc/systemd/system/hippotatd.service.d containing something like:

[Service]
User=my-hippotat-user

The init script checks that USER (possibly set by
/etc/default/hippotatd) and LOGNAME (set by systemd based on User=)
agree to make sure that both sources agree on what user hippotatd should
run as.

Merge branch 'sd-notify' into 'main'

Use systemd notify protocol to signal startup, when requested in env

See merge request iwj/hippotat!58

Merge branch 'update-lockfile' into 'main'

Routine dependency updates

See merge request iwj/hippotat!59

Document use of systemd notify protocol, in the usage message

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Use systemd notify protocol to signal startup, when requested in env

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Add sd-notify dependency

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Allow use of nix 0.31 (and use in upstream Cargo.lock)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Update upstream deps (Cargo.lock)

Route `cargo update`

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'update' into 'main'

Add more fixes to 1.2.3, and prepare for full release

See merge request iwj/hippotat!56

changelog: Document the remaining changes for 1.2.3

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Set cargo package versions to 1.2.3

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

debian/control: Drop mention of nonexistent hippotat-doc package

The docs are in -common.

Closes: #1115401
Closes: #1115402
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'changelog' into 'main'

Document changes in 1.2.3 (planned Debian-only release)

See merge request iwj/hippotat!55

Document changes in 1.2.3 (planned Debian-only release)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'nix-30-redux' into 'main'

Compatibility with nix 30, redux

See merge request iwj/hippotat!54

Allow use of nix 30

No change to lockfile (used by upstream-style build system) for now.

Closes: #1116421
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Use libc::dup2 and libc::read rather than nix versions

nix is doing io-safety which is an incompatible change.  It's hard to
write code which works with both versions of nix because BorrowedFd is
in Rust 1.66 which is higher than our MSRV.

Partially based on code from Peter Green.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Use .into_raw_fd for null_fd for compatibility with nix 30

In nix <=29, open returns RawFd for which .into_raw_fd is a no-op.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'deftly' into 'main'

Switch to derive-deftly, and release 1.2.2

See merge request iwj/hippotat!52

Update version to 1.2.2

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

macros: Allow use of syn 2

Now there is almost nothing here, we don't need any code changes
between syn 1 and syn 2.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

macros: Drop unneeded dependency on itertools

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config: Fix some misspellings

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config: Remove manual entry for inspecting `link`

InspectableConfigAuto handles this now, since
  config derive: Switch to derive-deftly

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config derive: Improve docs

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config derive: Have special no longer take an argument

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config derive: Improve and properly document InstanceConfigGlobal

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config derive: Use a define

This abolishes the helper type.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

config derive: Switch to derive-deftly

Replace our bespoke proc_macro with a derive-deftly macro.

The macro is based on one written for derive-adhoc 0.0.1, so could do
with a bit of updating.

I've inspeccted the changes to the generated output with a semi-manual
diff.  The only significant changes were changes to
`impl InspectableConfigAuto for InstanceConfig`, as follows:

>    +                    "link" => &self.link,

This is case is handled by the manually-written
`impl InspectableConfig for InstanceConfig` so never reaches
the derived impl.  But it is identical.

>    +                    "max_batch_up" => &self.max_batch_up,

This seems to have been an omission in the previous code.
So the addition seems good.

>                         "max_clock_skew" => &self.max_clock_skew,
>    -                    "max_clock_skew" => &self.max_clock_skew,
>    -                    "ifname_server" => &self.ifname_server,
>                         "ifname_server" => &self.ifname_server,

This is the removal of two duplicates, which arose from the odd code
structure in the previous code, which does code gen for inspect_key
once per attribute, rather than once per field.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update (minimal)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Routine cargo update

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Makefile: Force pipe-based jobserver

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'release' into 'main'

Release 1.2.1

See merge request iwj/hippotat!51

Release 1.2.1

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Merge branch 'fixes' into 'main'

Fix the autopkgtests in ci.debian.net

See merge request iwj/hippotat!49

Merge branch 'update' into 'main'

Routine Cargo.lock update

See merge request iwj/hippotat!50

autopkgtest: adt-initscript: use restart to start

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtest: adt-initscript: require rsyslog init script

We run rsyslog, so we do need its init script (unless we're running
under systemd).

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtest: adt-initscript: turn on set -x in the init script

Via /etc/default.  This may help debugging.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtest: adt-full: do not start a duplicate uservd

This is already done as part of setup-pair by setup-host.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

autopkgtest: adt-full: allow systemd-sysv as alt to init script

orphan-sysvinit-scripts and systemd conflict.  The effect is that in
ci.debian.net this dep would want to change the init system.

What we actually need is rsyslog to be started, which is satisfied
with systemd-sysv since then systemd is the init system and rsyslog
has a systemd unit.  (I believe.)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>