until it's been allocated.

This patch delays the VXID allocation until the time we know that we
will actually be logging anything for this request (not completely all
whitespace noise). This keeps the code in one place only (plus the ESI
mock request handling), instead of separated between the linger wait,
cache acceptor and pipelining code paths.

Coincidentally fixes the logging of timestamps problem in #1588.

Test case by daghf.

Fixes: #1581 #1588

This enables catching batch related log problems using logexpect.

This commit will make several test cases start failing until #1588 has been fixed.

healthy backend.

Fixes: #1575

will also use it.

For master we retain the more strict check in case we forget to bump
the API version as discussed on irc.  Implementation idea from scoof.

Fixes #1538

Not being able to fchown() is normal when running varnishd
unprivileged, so don't report failure unless running as root.

Add missing braces, avoid unnecessary variables.

Noticed by:	fgs

Extensive test cases often lead to a lot of log output,
which at times tripped the following assert:

Assert error in vtc_log_emit(), vtc_log.c line 122:
  Condition(vtclog_left > l) not true.

Double the amount of buffer space allowed to avoid this. Another 256KB
memory use per concurrent running test should be fine.

a VFP function but rather a fetch function calling VFP.

hackishly used.

Requests that end up in the hard "400 Bad Request" handling used to be
logged with incomplete data. (no method, maybe no path, possibly no proto,
and no response status)

Port scans or anything sending a byte and a linefeed would be logged.

Since this is used for logging access to a web site, it makes more sense
to skip these garbage requests.

Fixes:		#1584

Adding a dash if missing the path component does not make sense.

Related bug:	#1584

child process during startup.

Fixes #1407

This prevents SLT_ReqAcct log records with 0 as VXID.