Spotted by:	@jambon69

or replace the code wholesale, whichever comes first.

Most of the work was done in a93c4555, this
commit fixes remaining regressions compared with 5.1.2 release

Fixes #2340

... which were added in the meantime

The problem must be elsewhere, vtest results show no improvement

This reverts commit 57a3042d.

as canary on the API side.

not sure if they improve the situation, just checking

.map file anyway.

program will not need to map in all the VSC segments.

Also add back the "formats" property for VSC counters.

This is the only place where a `struct vbc *` pointer isn't called `vbc`
and this helps me get a better picture.

MacOS will return EINVAL under e.g. setsockopt if the connection was
reset.

Return "varnishd" for server.identity if no -i argument is given.

Wrap setproctitle() in mgt_util.c

But it should be available via VSM.

Otherwise the backend might see multiple Connection entries.
Fixes #2337.

If txrst is used without -err we want a valid error.
Spotted by msan.

Spotted by ubsan.

c00037.vtc already tests hash_always_miss.

In the section related to make test, the sentences were a bit hard incorrect. I've corrected it.