At least one case was brked.

Submitted by:	Federico G. Schwindt

Submitted by:	Federico G. Schwindt

Now only 100 more to go.

the actual error messages are somewhat less than helpful.

badvcl in that you can make sure it fails for the right reason.

All -badvcl's should be converted to -errvcl without undue delay.

Clean up some asserts.

and more orthogonal.

This is a "it works!" snapshot, there are improvements to the generated
C-code in the pipeline.

Use VSL_Dispatch instead of VSL_NextSLT in varnishtop, which does this
automatically for us.  Clean up accumulate() a bit in the process.

Fixes #1213

VCL reference it might have, in order to not hold on to discarded VCLs
forever.

types.  Have the python code learn the mapping from them.

Various cleanup and tightening of code.

set req.http.Cookie = req.http.Cookie + " " + req.http.Cookie-sess;

do the obvious thing, even if req.http.Cookie does not exist.

The underlying issue was a badly though through overloading of
the NULL value to mean "Unset".  Now it has its own magic marker.

Fixes #1218

Be more aggressive about tracking const values and merge more const
strings at compile time.

the goal-posts a serious distance while we're at it:

We can use the thread workspace for small temporary allocations during
VCL execution, use this to allocate a compound type for the HEADER
type, to get us closer to typedef-ability for VCL mapped types.

Ideally we could just reset it, but other in-fsm-state uses of this
workspace is allowed, so we cannot know for sure that the workspace
is empty when we begin.

to keep it light and small, so we forbid any long-term usage an detect
leakage with this assert.

Fixes	#1206
Found & Fixed by:	DocWilco

Found & Fixed by:	Federico G. Schwindt
Fixes #1212

0ad568 should never have been pushed.

This reverts commit 4f6eaa24.