Build fails with gcc 7.3 (-Werror=implicit-fallthrough)

[marmarek]

Build fails with this error:

+ /home/user/.opam/mirage-ssh-agent/bin/ocamlfind ocamlc -g -ccopt '--std=c99' -ccopt -Wall -ccopt -Wextra -ccopt -Werror -ccopt -O3 -ccopt -DACCELERATE -ccopt -msse2 -ccopt -maes -package zarith -package sexplib -package ppx_sexp_conv -package cstruct -c src/native/aes/aesni.c
src/native/aes/aesni.c: In function ‘_nc_aesni_enc_blocks’:
src/native/aes/aesni.c:316:19: error: this statement may fall through [-Werror=implicit-fallthrough=]
   __blocked_loop (_nc_aesni_enc, _nc_aesni_enc8, src, dst, rk, rounds, blocks);
                   ^
src/native/aes/aesni.c:294:9: note: in definition of macro ‘__blocked_loop’
         f1 (__b (src, 6), __b (dst, 6), rk, rounds);         \
         ^~
src/native/aes/aesni.c:295:7: note: here
       case 6:                                                \
       ^
src/native/aes/aesni.c:316:3: note: in expansion of macro ‘__blocked_loop’
   __blocked_loop (_nc_aesni_enc, _nc_aesni_enc8, src, dst, rk, rounds, blocks);
   ^~~~~~~~~~~~~~
(...)
cc1: all warnings being treated as errors
Command exited with code 2.

The environment is Fedora 26, ocaml 4.05.0

[pqwy]

Hmm it seems that this bit never made it into a release.

Ok, there should be a release sooner rather than later.

[marmarek]

IMHO it is useful warning to have. But feel free to close this PR if you want it disabled globally anyway.

[cfcs]

@pqwy A higher upper bound on cstruct would be nice in a future release too! (Since Cstruct has added/changed some new stuff over the past 13 months)

[pqwy]

@cfcs Sure, it's a quick stop-gap measure since cstruct gained homegrown element access I'm not completely sure about.