Spotted by:	gquintar

impatience to make -t argument controlling

doc fixes #2107

This test wasn't working as it was supposed to, and would cause the
forward links graph building of log transactions to fail. The fail was
soft, as the reverse mapping still succeeded creating the right
mapping.

Depending on when the local thread log buffer was flushed, the problem
would show as spurious "duplicate link" log records to be inserted.

Fixes: #1830

Fixes #2105. Minor test cleanups while here.

Let's see how many grammar/spelling fixes are going to follow :/

More to come.

Fixes #2017.

This reverts commit 6da0cfda.

That was a stupid commit, the malloc failure is already handled

Spotted by: @scn

Since a78efad8 the argument name and default
value are separated from the argument spec by a \1, so we need to stop
there.

Fixes: #2094
Fixes: #2096

makes a valid H1 request header.

One submitted by github::bobmshannon.

The VARNISH_TEST_PATH variable contains the minimal PATH for running a
test suite, covering only Varnish tools. And VARNISH_LIBRARY_PATH is
needed because Varnish binaries aren't build using an RPATH and only
know about their libraries out of the box when they are installed in
the system path, via ldconfig.

Consequently VMOD_TEST_PATH is now an alias for VARNISH_TEST_PATH.

Catching this problem after the 5.0 release is unfortunate, as it won't
be ported to 4.1 to avoid breaking the upgrade path between the 4.1
series and the 5.0.0 release. Unless it is considered OK?

Do people really read through the whole commit message and diff when it
concerns autotools changes?

While at it, make a pass on the autoconf documentation. Flatten code
snippets that don't really need line breaks. Add the man page to the
dist tarball by default. Walk through a typical test suite setup.
Document variables created or substituted by AC_PREREQ. Document how
the purpose and requirements of varnish.m4, and document what changes
between versions of Varnish.

Consequently bump the serial number.

Polish test while here.

The long description for MAIN.cache_hit in varnishstat looked
silly with the double/triple spaces.