implemented from scratch, an optimized variant of strcpy_s.
Note that the intel fork also has it, with the same API.

This adds a src termination check, when src is static.

not as far as travis qemu/docker builds yet
not much archs

fixed mingw autoconf by setting the crlf policy. see previous commit.

esp. hardcode $as_echo as echo.
But the real problem is crlf conversion on multiple \ lines, leading
to \^M on mingw.

no CONFIG_COMPAT_32BIT_TIME
Should fix GH #96

and remove not generated wcsnaticmp_s.3 manpage. Fixes GH #95

the tardist name went from libsafec-02092020 to safeclib-3.7.
this should stay the established public name.
packagers still should use libsafec and libsafec-devel for the lib parts.

without git tag versioning for now.
See GH #91

fixup some MD markups

which should help rst

aarch64 has no mrc opcode.

and use it.
p_memset_s needs to link to mem_primitives_lib.lo for windows.
it is not exported.

as in smhasher.
also change the percent to compare against. mem_prim_set vs memset

gcc now as fast as memset, only clang 2x slower without -march=native.
With -march=native gcc is 2.5% faster than glibc, clang 3% slower.

to test the MEMORY_BARRIER overhead and compare the fast mem_prim_set
with native memset.

gcc and clang overheads now both bad: 56/51% slower.
apparently glibc got faster in between.

They are already in all memset/zero/cpy calls. we dont need it twice, keep
mem_prim_set fast

no idea why this fails.
use the debian image, as with travis.

* duplicate return values.

Skip the _chk functions, only document the public macro API.
See GH #33

ubuntu might be too broken, and I cannot debug it.
it works on debian, fedora, centos and rhel.

from https://github.com/junaruga/ci-multi-arch-native-container-test
which do work with docker, but not podman.
for arm64 and s390x.

TODO failing images:
* memcmp16_s fails with ubuntu-aarch64-linux-native,
  but passes {debian,centos,fedora}-aarch64-linux
* gmtime_s, localtime_s fails ubuntu-arm32-linux-qemu (arm32v7),
  ubuntu-s390x-linux-qemu
  but passes: debian-arm32-linux, debian-s390x-linux

generally ubuntu looks like trash. wonder what causes these fails?
custom glibc patches? unfixed bugs?

rename build-aux/conf-make-check.sh

s390x, ppc64le, arm64

See GH #85

ditto for ranlib and nm

See GH #85. Annotate our decls accordingly.

See GH #89

use old_time32_t then for our API.
We didnt add the new time64_t APIs yet