Replace explicit recursion by heap allocations
in sexp_mark_one code.
This prevents crashes caused by stack overflow.
In particular, fixes issue #601.
As an optimization, allocate a fixed sized stack buffer first,
which should be enough for "normal" uses.
When that stack overflows, switch to heap.
Also, store "ranges" on the stack, instead of the actual sexp's,
using the fact that sexp's of a single parent are continous in memory.
This patch doesn't remove recursion on the context saves
because it didn't seem like they overflow in practice.
But changing that is simple having the stack interface.
On NetBSD systems the netbsd feature identifier was not actually
provided because the macro SEXP_NETBSD was defined twice and the
second value was 0.
On DragonFly systems the dragonfly feature identifier was not actually
provided because the macro SEXP_DRAGONFLY was defined twice and the
second value was 0. Also on DragonFly systems the darwin feature
identifier was incorrectly provided because SEXP_DARWIN had a value of
1 instead of 0.
Introduce SEXP_STATIC_LIBRARY to support static-library build on
Win32/Win64. On Win32, symbol visibility is enforced on DLL builds so we
need to dedicated ABI on static-library builds.
Under Unix with SEXP_64_BIT defined, sexp_sint_t is defined as 'long'.
But we would get the equivalent format specifier SEXP_PRIdFIXNUM from
the OS-defined PRId64 in <inttypes.h>. MacOS defines it as "lld". This
causes the clang printf checker to emit a warning about the 'long' and
'long long' mismatch.
Fix by avoiding system-defined PRId32 and PRId64 format specifiers and
always defining SEXP_PRIdFIXNUM as "d", "ld" or "lld" according to our
definition of sexp_sint_t as int, long or long long. This also means
we don't need to include <inttypes.h> any more.
sexp_double_to_ratio_2, which converts without introducing
round-off errors the way sexp_double_to_ratio does when it
multiplies by 10.
Changed sexp_inexact_to_exact to use this new function when
a non-zero fractional part of the input exists.
