Re: [Tarantool-patches] [PATCH luajit] test: add test for conversions folding

[Sergey Bronnikov via Tarantool-patches <tarantool-patches@dev.tarantool.org>]

Hello, Sergey!

LGTM

On 11/16/23 16:44, Sergey Kaplun wrote:
> This patch adds the test for commit
> 1a401622fe83ae695226c94df3416e312315d659 ("Fix assertions."). This patch
> removes incorrect assertions in the fold optimizations for conversions
> from numbers to different integer types. Although the issue affects only
> branch 2.0, there is no need to fix it. Nevertheless, the test is
> required to avoid regressions in the future.
>
> Part of tarantool/tarantool#9145
> ---
>
> Branch: https://github.com/tarantool/luajit/tree/skaplun/lj-833-fold-conv-from-num
> Tarantool PR: https://github.com/tarantool/tarantool/pull/9375
> Related issues:
> * https://github.com/LuaJIT/LuaJIT/issues/833
> * https://github.com/tarantool/tarantool/issues/9145
>
>   .../lj-833-fold-conv-from-num.test.lua        | 33 +++++++++++++++++++
>   1 file changed, 33 insertions(+)
>   create mode 100644 test/tarantool-tests/lj-833-fold-conv-from-num.test.lua
>
> diff --git a/test/tarantool-tests/lj-833-fold-conv-from-num.test.lua b/test/tarantool-tests/lj-833-fold-conv-from-num.test.lua
> new file mode 100644
> index 00000000..9e2059bd
> --- /dev/null
> +++ b/test/tarantool-tests/lj-833-fold-conv-from-num.test.lua
> @@ -0,0 +1,33 @@
> +local tap = require('tap')
> +
> +-- XXX: Test the behaviour of fold optimizations from numbers to
> +-- different integer types. The test itself doesn't fail before
> +-- the commit since these changes relate only to version 2.0.
> +
> +local test = tap.test('lj-833-fold-conv-from-num'):skipcond({
> +  ['Test requires JIT enabled'] = not jit.status(),
> +})
> +
> +local ffi = require('ffi')
> +
> +test:plan(3)
> +
> +local arr_i64 = ffi.new('int64_t  [2]')
> +local arr_u64 = ffi.new('uint64_t [2]')
> +local arr_u32 = ffi.new('uint32_t [2]')
> +
> +jit.opt.start('hotloop=1')
> +
> +for _ = 1, 4 do
> +  -- Test conversion to type (at store). Also, check the
> +  -- conversion from number to int64_t at C array indexing.
> +  arr_i64[1.1] = 1.1
> +  arr_u64[1.1] = 1.1
> +  arr_u32[1.1] = 1.1
> +end
> +
> +test:is(arr_i64[1], 1LL,  'correct conversion to int64_t')
> +test:is(arr_u64[1], 1ULL, 'correct conversion to uint64_t')
> +test:is(arr_u32[1], 1ULL, 'correct conversion to uint32_t')
> +
> +test:done(true)