pkg_libhydrogen tests fail / update libhydrogen

[chrysn]

Description

The libhydrogen tests fail with scary errors on recent GCC:

/home/chrysn/git/RIOT/build/pkg/libhydrogen/impl/sign.h:104:5: error: ‘hydro_x25519_core’ accessing 160 bytes in a region of size 32 [-Werror=stringop-overflow=]
  104 |     hydro_x25519_core(&xs[2], sig, hydro_x25519_BASE_POINT, 0);
      |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/chrysn/git/RIOT/build/pkg/libhydrogen/impl/sign.h:104:5: note: referencing argument 1 of type ‘hydro_x25519_limb_t[5][8]’
 {aka ‘unsigned int[5][8]’}
/home/chrysn/git/RIOT/build/pkg/libhydrogen/impl/sign.h:104:5: note: referencing argument 2 of type ‘const uint8_t[32]’ {aka ‘const unsigned char[32]’}
/home/chrysn/git/RIOT/build/pkg/libhydrogen/impl/x25519.h:251:1: note: in a call to function ‘hydro_x25519_core’
  251 | hydro_x25519_core(hydro_x25519_fe xs[5], const uint8_t scalar[hydro_x25519_BYTES],
      | ^~~~~~~~~~~~~~~~~

Steps to reproduce the issue

• Use recent GCC
• make -C tests/pkg_libhydrogen all

Issue references

This is fixed upstream in jedisct1/libhydrogen#123 -- and the patch that closes it indicates that it's "purely cosmetic" (ie. ignoring the errors, as an old compiler would do it, does no harm).

Backport status

will need to be evaluated based on the amount of upstream changes; if excessive, a patch could be backported (but that'd need to happen before the update hits the master branch). For now I plan to list that as known issues; tagging as "needs backport" still to keep track.

[maribu]

fe428c3 fixed this