Changes

Drop the ad-hoc qt6 build deps from the deb images

Update library documentation job

Add the missing omc-with-jq

Updated nightly build

Install the nightly omc for the documentation

The `openmodelica/openmodelica:v1.27.1` image does not exist, and 1.27.1
is too old to parse the libraries (SDF) anyway. Unpack the nightly CLI
tarball into the workspace instead; it is relocatable and needs only
glibc, libstdc++ and the CA bundle from the image, which is now a plain
ubuntu that stays in the docker build cache.

Clean the whole workspace: the cargo target tree kept between runs grows
without bound.

Assisted-by: Claude Opus 5 (1M context)

Updated nightly build

Updated nightly build

Updated nightly build

Bound the documentation job with a cgroup

`--max-memory` was a `RLIMIT_DATA` ceiling, which counts address
space. jemalloc retains freed extents rather than unmapping them, so
that figure only grows while resident memory falls back, and the run
was killed at 96 GB while holding 55 GB. The container now carries a
`--memory` limit, which is charged for resident pages, and
`--max-memory` is the live-memory budget the generator works to --
half the cgroup, since retained extents and fragmentation sit on top
of it.

Both figures and the job count now come from what the container
actually has, `nproc` and `memory.max`, rather than being pinned to
what the node label promises. Capped at ten threads: past that the
allocator's retained address space grows far faster than the work
does, and rendering is bound by memory traffic anyway.

Assisted-by: Claude Opus 5 (1M context)

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Pass --no-lintian before the dpkg options

debuild stops parsing its own options at the first argument it does
not recognise and hands the rest to dpkg-buildpackage. `-d` is such
an argument, so --no-lintian was forwarded and rejected:

    dpkg-buildpackage: error: unknown option or argument --no-lintian

debuild turned that into a `die`, which exits with errno when it is
non-zero - ENOTTY (25) on a Jenkins agent with no tty - so the
nightly failed with the fairly opaque `exit status 25`.

Assisted-by: Claude Opus 5 (1M context)

Updated nightly build

Updated nightly build

Speed up the source package build

lintian accounted for more than half the wall clock of a
Jenkinsfile.sourcepackage run: for a source package it unpacks the
whole openmodelica orig tarball again and scans every file, once per
built revision. Nothing consumes its output - it is neither archived
nor gated on - so drop it with debuild --no-lintian. That also removes
a failure mode where a new lintian version invents an error tag and
fails the nightly.

Compress the orig tarball with `xz -T0` instead of tar's single
threaded `cJf`. Same preset, so the tarball only grows a fraction of a
percent from the block splitting, and it decompresses in parallel too.
Note this wants roughly 100 MB of encoder memory per core; pin the
thread count if that ever gets tight on the builder.

Assisted-by: Claude Opus 5 (1M context)

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build

Updated nightly build