testsuite fails on arm{el,hf} and ppc64el

[mehdid]

This bug is forwarded from Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=907431

The testsuite of cppo fails on arm{el,hf} and ppc64el in the same location:

https://buildd.debian.org/status/fetch.php?pkg=cppo&arch=ppc64el&ver=1.6.4-1&stamp=1533482737&raw=0
https://buildd.debian.org/status/fetch.php?pkg=cppo&arch=armhf&ver=1.6.4-1&stamp=1533483827&raw=0
https://buildd.debian.org/status/fetch.php?pkg=cppo&arch=armel&ver=1.6.4-1&stamp=1533484809&raw=0

dh_auto_test -a
make -j4 test
make[1]: Entering directory '/<>'
cppo alias test/runtest (exit 1)
(cd _build/default/test && ../../install/default/bin/cppo test.cppo) > /dev/null
Error: File "test.cppo", line 98, characters 0-20
Error: math error
make[1]: *** [Makefile:5: test] Error 1
make[1]: Leaving directory '/<>'
dh_auto_test: make -j4 test returned exit code 2
make: *** [debian/rules:18: build-arch] Error 2

[pirbo]

The culprit is the test -1 asr 100 = -1. It reveals 2 problems:
An obvious mistake line 255 of src/cppo_parser.mly where:

  | aexpr ASR aexpr          { `Lsr ($1, $3) }

should be

  | aexpr ASR aexpr          { `Asr ($1, $3) }

and a potentially dangerous behaviour lines 224 - 252 of src/cppo_eval.ml:
The code ensures that -64L <= shift <= 64L but the OCaml documentation says [...] The result is unspecified if y < 0 or y >= 64 so you might want to ensure 0L <= shift < 64L...

[mjambon]

Thanks @mehdid and @pirbo. 1.6.5 was released. The opam release is in progress (ocaml/opam-repository#12606).