size of the request, probably sleep copy&paste ?

Fixed	#1175
Submitted by:	martin

It was a nice idea, but obviously compiler-writers and/or language
laywers are pussies who hate table-based programming :-)

which subprocess we are locking down.

Inspired by Geoff's patch for Solaris, but does not solve the
Solaris problem, only makes it easier to fix.

Fixes a compile error (incorrect signature for SES_Delete() in
cache_waiter_ports.c)

'param.show -l'.

Fixes: #1169

If the ban lurker finds a busy object, skip the object and defer the
checking to a later pass.

No test case because it is very timing sensitive (but see the trac bug
report for a VTC-file that mostly triggers it on the 3.0 branch only)

Fixes: #1162

There is a short window of opportunity between HSH_Insert() and
HSH_Unbusy(), where the locks are released and the expiry thread can
try to expire an object that is still busy (for objects with very
short TTL). This patch makes the expiry thread check for busy objects
and wait if the head of the binheap is busy.

No test case as the window is too small.

Fixes: #1150

An attempt to nuke a busy object will end in assertion.

eliminate two SES_Charge() calls that have nothing to charge.

when we come back from the waiter, to lessen amount of work
waiter does.

Fixes: #1164

C-standardisation.  (Even money if you ask me...)

GCC users will just have to live with marginally works docs.

when using the -n option.

Code had reversed the no -T argument error check, and also required
the optional -S argument to be present.

Fixes: #1160

Some breakage on the far side of varnishapi expected, mode to come.

VSL record for a session.

Make $log_local_address default to on.

in the future.

unique transaction id 'vxid'

by moving the VSL'ing of socket endpoints earlier in the processing.

NB: Changes to parameter defaults & units included in this commit!

I tried to apply some queuing theory to this problem and as much as
I admire Agner Krarup Erlang, his math isn't much help when you don't
know any of the relevant metrics for your queue.

Instead I took a much simpler approach:  "If we fail to get a thread,
we probably need more threads", and have rewritten the herder to to
react faster and more reliably to such events.

I went over the parameters for thread-pools and normalized timeouts
to seconds rather than milliseconds (beware!) and polished descriptions
etc.