from the log:

**   top   === shell -exit 1 -expect {failing as requested} {
**** top   shell_cmd|exec 2>&1 ;
**** top   shell_cmd|\tvarnishadm -n /root/VT/_vtest_tmp/vtc.37277.0dc3c11b/v1 vcl.load f1 /root/VT/_vtest_tmp/vtc.37277.0dc3c11b/f1
**** dT    27.789
**** v1    vsl|          0 CLI             - Rd vcl.load f1 vcl_f1.1686893821.277415/vgc.so 1auto
**** v1    vsl|          0 CLI             - Wr 300 62 VCL "f1" Failed initialization
Message:
	failing as requested

**** dT    27.893
**** v1    vsl|          0 CLI             - Rd ping
**** v1    vsl|          0 CLI             - Wr 200 19 PONG 1686893826 1.0
**** dT    32.987
**** top   shell_out|CLI communication error (hdr)
**** top   shell_status = 0x0002
---- top   shell_exit not as expected: got 0x0002 wanted 0x0001
*    top   RESETTING after ../../../../bin/varnishtest/tests/m00000.vtc
**** dT    32.988

in the most diff-friendly way.

If Mark Adler ever decides that the 21st century is real thing, we
will adopt his diff.

This should allow us to revert 79c7d175

Varnish historians claim to have evidence that the odd wording "from
mgt_param" could have been an unintended side effect of a struct
rename in ancient times at around 2011. (82411494)

To keep things simple, the open delay is global - the delay of
the last debug stevedore applies to all of them.

With the recent change to stream the request body to the backend like
chunked encoding does, even when we know the content length, two test
cases became sensitive to the timing of window updates. The data frames
contain for the most part a 16kB payload, which happens to be the same
as fetch_chunksize's default value.

With a value of 64kB, corresponding to one more byte than the default h2
initial window, we are garanteed to never risk a race between the VFP
consuming the h2_rxbuf and the h2_sess task monitoring it to send window
updates.

This is also a good occasion to exercise stream.peer_window in t02019.

Refs #3878
Refs #3934

Seen while staring at #3937

Format individual subroutines as code, add a trailing period.

This was meant to be tested by 054215b5,
but the test did not attempt to receive a second chunk, so it succeeded
when it should not have.

Fixes #3938

When VCC is run, some temporary files are generated during the
process and are cleaned up at the end. The cleanup tries to remove
all the files that are supposed to be generated and displays a
"Could not delete 'vcl_...': No such file or directory" message if it
fails to find one of them. The error happens when VCC fails before
generating the file(s). This commit avoids the bogus messages by
ignoring ENOENT if VCC failed.

Fixes #3925

It's about time...

The --enable-maintainer-mode option has been used in autogen.des for a
while now and it has allowed automatic reconfiguration and hot reload of
makefiles.

For the next step, we can turn certain programs into maintainer programs
and only require them when they are absolutely needed. This is what we
currently do with SVG files generated from dot files, checked in git and
optionally required, with an error message when rebuild is not possible.

It turns out we only need 3 lines in configure.ac to implement all that,
with the following differences:

- no more `./configure --with-dot[=...]` option
  - use `./configure DOT=...` to override the default `dot` command
  - or alternatively run configure with `DOT=...` in the environment
- generic missing error message (but referencing dot) from make
- no need to reconfigure once `dot` is in the PATH

The DOT override is actually documented in `./configure --help` like the
`--with-dot[=...]` option was.

While dot is our most emblematic maintainer tool, with its build targets
checked in all the way in the git repository, this could be applied to
other maintainer programs like sphinx-build and rst2man that should not
be required when building from a dist archive.

We don't need to render witness.svg in CI, even if dot(1) is available.

Should we give it an explicit ETIMEDOUT in the dripping write case?

Better diff with the --word-diff --word-diff-regex=. options.

This is to ensure we don't lose write errors by the time we make the
final checks.

Instead of trying to allocate storage for the entire request body when
the content length is known, let it go in fetch_chunksize steps. This
may prevent spurious cache evictions for large request bodies for pass
transactions when we are merely streaming the body.

On the other hand, when the content length is known and the goal is to
cache the request body, we may still attempt a single allocation and all
that entails.

Also motivated by #3935

motivated by #3935

https://editorconfig.org/

Spotted by Coverity Scan.

Refs #3914

Refs #3929

With that we lose the ability to rebuild the manual from the doc/sphinx
directory if the sources were modified, just like any other dependency
on $(top_builddir) sources. This one is linked once and for all, but
doesn't fit the link_srcdir target.

Suggested by @nigoroll.

This guarantees that VSC code is generated before compiling anything.

Fixes #3928