make it work more like people would expect.

Fixes #1991

die until we have started the child process.

what our job is and how we will structure processes for it, before
doing all the complex stuff.

startup situation.

Example for the case addressed here is an html comment which is not esi.

	1st call on !--e
	2nd call on !--end

The first call could match !--esi:, so vep_match() does not return with
VEP_NOTMYTAG, but with null, so characters up until 'e' are marked.

For the second call, we need to consider the case that characters after
the end of the tag could already be marked.

Fixes #1988

ESI Debugging code is disabled unless the Debug macro at the top of the file is changed.

it is the place which defines
VMB_NEEDS_PTHREAD_WORKAROUND_THIS_IS_BAD_FOR_PERFORMANCE

cache.h before system includes to avoid duplicate include warnings
- except for stddef.h, which should come before cache.h to avoid warnings on
  Linux/gcc
some reordering for the same reason
remove some unused includes
misc cleanup

Still incomplete

Info 793: ANSI/ISO limit of 31 'significant characters in an external
    identifier', '__pthread_unregister_cancel_restore(struct
    __pthread_unwind_buf_t *)', exceeded -- processing is unaffected
Info 793: ANSI/ISO limit of 31 'significant characters in an external
    identifier', 'pthread_mutexattr_getprioceiling(const union
    pthread_mutexattr_t *, int *)', exceeded -- processing is unaffected
Info 793: ANSI/ISO limit of 31 'significant characters in an external
    identifier', 'pthread_mutexattr_setprioceiling(union pthread_mutexattr_t *,
    int)', exceeded -- processing is unaffected

duplicate includes via cache.h and common.h

export missing VSB symbols from libvarnishapi

Don't:

	MACRO(
		a,
	#if FOO
		b,
	#else
		c,
	#endif
		d
	)

A function like macro was invoked whose arguments extended for
multiple lines, which included preprocessor statements. This is
almost certainly an error brought about by a missing right parenthesis.

By the rules of Standard C the preprocessing directive is absorbed
into the macro argument but then will not subsequently get executed.
For this reason some compilers treat the apparent preprocessor
directive as a directive. This is logical but not portable.

It is therefore best to avoid this construct.

wouldn't make sense.

Fixes: #2099

NB: This _may_ expose certain corner cases of unwise symbol overloading
    in VCL.

no revision 1.7

thanks @dridi

am__check_pre in the automake-generated Makefile basically runs

	test -d tests || mkdir tests || exit $?

When running out-of-tree with parallel make (as vtest does), this
becomes racy. Avoid the race by a mkdir ignoring the error

The previous commit did not solve the issue for all cases
- we also change OFLAGS in configure

In particular with --enable-developer-warnings, we select strict compiler
options for CFLAGS, which are too strict for the configure built-in conftest.c
code.

Make distcheck calls configure again for an out-of-tree build, so it would
see the CFLAGS the "outer" configure built. To avoid this situation, we
restore the original CFLAGS for the inner configure.

I wonder if there is a better, more autoconf-ish, solution to this problem.

Tested with SolarisStudio 12.4 on amd64/x86