it at the end of the run where it is far more likely to be noticed.

This *mainly* affects VSB_QUOTE_NONL.

Make sure VSB_QUOTE_NONL always ends with a NL

of the while { } do(0) in AZ...

function to use instead of vtc_log(..., 0,) in trivial cases.

It looks like it was simply a difference between automake versions,
let's see if it pans out.

Refs #2139

A failed fetch won't have the storage trimmed, which may result in an
empty storage element in the body list. Do not assert on non-empty
size in the simple iterator.

Fixes: #2186

This can be reintroduced at a later time when someone feels like
fixing / rewritting it.

Related to #2147.

closes #2014

This round of fixes was originally from work-in-progress in #2174 but
the way forward is to remove the autotools. Since #2174 became obsolete
they should be checked in independently.

Those are warnings we should be able to comply to, but are commented
out since they are clang-specific, but I'm not sure how `optreset`
should be handled so I left that one alone.

Closes #2174

Refs #2139

not accidentally produce '--' and '++' operators.

Fixes	#2184

I noticed that some shells like ye olde bash available on SmartOS don't
treat the following construct as single a token:

    ${dir}/${file}

The following expands a single token on the RHS even if the variable
contains spaces:

    var=${dir}

But a token involving a regular string too needs to be protected by
double quotes:

    var="${dir}/${file}"

Even when none of the variables contain blanks.

Refs #2139

IOW if we ended up in vcl_backend_error{} or failed after returning
delivery in vcl_backend_response{} (e.g. we did not receive enough bytes)
we will bump this counter.

it slows things down.