It reached EOL almost 2 years ago, but the architectures we cover in CI
had 2 extra years of long term support that is coming to an end this
month.

With CentOS reaching EOL at the end of the month and CentOS Stream 9
flaky packaging jobs, we don't need to keep them around. For el9 builds
we rely on the almalinux:9 image.

We don't need to dial it up to 11 for the case where `make clean` runs
in a source tree coming from a release archive instead of a git clone.
Running `make distclean` should be enough to clear build artifacts that
may not be known by older branches before switching to them.

Otherwise we can't rely on the body status to make decisions when more
than one request is sent to an HTTP/1 connection.

It was instrumental in the understanding of a filter bug that was
previously difficult to troubleshoot.

It needs the debug::processors flag and Debug records to be unmasked to
become useful.

And finally, the test logs contain the complete VSL output for all four
transactions. It was an attempt at following the VSL in this test case
that led me to finding where the VSM status was lacking or incorrect.

Since there are two processes with possibly different credentials, track
the ability to kill them at the vsm_set granularity. When a restart is
established, stop assuming the ability to send signals to the process.

If a live VSM advertises a process that yields something other than
ESRCH, assuming that the PID is not recycled too soon, we can conclude
that EPERM is as good as a successful kill.

Summarily:

- index hash change means restarted
- index new plus or minus means changed
- reopening the directory means restarted

This was probably a copy-pasta accident of the previous line. It cannot
be tested with an instrumented "varnish v1" type of varnishd instance,
because it's not supposed to ever stop the MGT process once started.

This reverts commit a3593245.

This reverts commit b845c278.

The liveness check is brought back with its deficiencies that will be
solved separately.

This is an attempt at speeding up this job that usually takes three time
as long as other distcheck jobs.

This reverts commit 328540c6 in the
lib/libvarnishapi/vsc.c file where no magic check could be added to
replace AN(vsm) occurrences.

We should probably look for more potential net loss of assertions from
that commit.

the actual issue was that strtoll(".456", ...) sets errno 22, so when
we fall back to strtod(), we need to clear it.

Ref 0ca8d177
Ref 0dfa3b8c

Fixes #4088

Yes, I did install FreeBSD, and it's been a nice refreshing no-frills experience
so far.

as of 14.1-RELEASE

include cleanup and same change as cccb8c1d

Ref 7392333a

It seems I wrongly assumed strtoll(".", ...) would return 0, so
bring back the fallack to float parsing similar to what we had
before 05e10eee.

Should fix regression on FreeBSD from 0dfa3b8c

Also clean up now unneeded include (thank you, Flexelint).

Fixes #4088

I promise, if this is still wrong, I will install FreeBSD ;)

Fixes #4088

This is not the case for example on Alpine Linux.

But it is not required when building with libunwind support.

This facility allows intermediate layers to set a prefix without
meddling with the provided format string.

This is done in a VSB to match what the panic code is currently doing.

When libunwind is used but fails to produce a back-trace, there is now
a fallback to the original libexecinfo-based implementation (suggested
by @asadsa92).

I don't understand why they didn't maintain the unversioned centos:stream
tag as a kind of latest tag after dropping the actual centos:latest tag.

The plain "stream" image has not been updated for a year, and instead of
being the equivalent of a "latest" tag it just appears to be frozen.

Since there isn't an automatic centos-stream upgrade path, let's target
version 9 explicitly for now.