x = [99999999999999999999999999, -(2 ** 53 - 2)];
for (let i = 0; i < 2; ++i) {
    print((x[i] | 0) / 3);
}

$ ./js-dbg-64-dm-armsim64-linux-x86_64-d8321d0b2c5b --fuzzing-safe --no-threads --ion-eager testcase.js
0
0

$ ./js-dbg-64-dm-armsim64-linux-x86_64-d8321d0b2c5b --fuzzing-safe --no-threads --baseline-eager --no-ion testcase.js
0
0.6666666666666666

Tested this on m-c rev d8321d0b2c5b.

My configure flags are:

AR=ar sh ./configure --enable-simulator=arm64 --enable-debug --enable-more-deterministic --with-ccache --enable-gczeal --enable-debug-symbols --disable-tests --disable-cranelift

python3 -u -m funfuzz.js.compile_shell -b "--enable-debug --enable-more-deterministic --enable-simulator=arm64" -r d8321d0b2c5b

Due to the simplicity of this bug, compare_jit is finding mostly dupes of this on ARM64.

autobisectjs shows this is probably related to the following changeset:

The first bad revision is:
changeset: https://hg.mozilla.org/mozilla-central/rev/7ed7b0b9f2dd
user: Nicolas B. Pierron
date: Mon May 27 16:36:26 2019 +0000
summary: Bug 1523568 - ARM64: Implement divisions by a constant. r=sstangl

Nicolas, is bug 1523568 a likely regressor?

(In reply to Gary Kwong [:gkw] [:nth10sd] from comment #1)

Nicolas, is bug 1523568 a likely regressor?

Yes.

https://hg.mozilla.org/mozilla-central/rev/09088ff4d2a7