From: Evgeniy Temirgaleev via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: "Sergey Kaplun" <skaplun@tarantool.org>
Cc: tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH luajit] FFI/MacOS: Fix calling convention for on-stack varargs.
Date: Tue, 30 Jun 2026 17:46:39 +0300 [thread overview]
Message-ID: <1782830799.474292866@f378.i.mail.ru> (raw)
In-Reply-To: <20260629131914.791223-1-skaplun@tarantool.org>
[-- Attachment #1: Type: text/plain, Size: 6504 bytes --]
Hi, Sergey! Thanks for the patch! LGTM
--
Best regards,
Evgeniy Temirgaleev
>
> From: Sergey Kaplun <skaplun@tarantool.org>
> To: Sergey Bronnikov <sergeyb@tarantool.org>, Evgeniy Temirgaleev <e.temirgaleev@tarantool.org
> >
> Cc: tarantool-patches@dev.tarantool.org, Sergey Kaplun <skaplun@tarantool.org
> >
> Date: Monday, June 29, 2026 4:19 PM +03:00
> From: Mike Pall <mike>
>
> Thanks to Sergey Kaplun.
>
> (cherry picked from commit a2bde60819d83e6f75130ac2c93ee4b3c7615800)
>
> This commit fixes the regression introduced by the commit
> 82ca6844234901366c043cd103f6eae345af753c ("FFI/MacOS: Fix calling
> convention for enums."). The `isva` flag is set to 0 even for vararg
> functions. Thus, arguments on the stack may be aligned incorrectly,
> leading to the crash.
>
> This patch fixes the behaviour by adjusting the flag value instead of
> resetting it. In addition to the original test, various tests have been
> added to cover the behaviour of the vararg FFI calls.
>
> Sergey Kaplun:
> * added the description and the test for the problem
>
> Part of tarantool/tarantool#12480
> ---
>
> Branch: https://github.com/tarantool/luajit/tree/skaplun/lj-1455-macos-arm64-vararg-regression
>
> Related issues:
> * https://github.com/LuaJIT/LuaJIT/issues/1455
> * https://github.com/neovim/neovim/pull/39994
> * https://github.com/tarantool/tarantool/issues/12480
>
> src/lj_ccall.c | 2 +-
> ...455-macos-arm64-vararg-regression.test.lua | 112 ++++++++++++++++++
> 2 files changed, 113 insertions(+), 1 deletion(-)
> create mode 100644
> test/tarantool-tests/lj-1455-macos-arm64-vararg-regression.test.lua
>
> diff --git a/src/lj_ccall.c b/src/lj_ccall.c
> index 7c3ec1e5..ed692318 100644
> --- a/src/lj_ccall.c
> +++ b/src/lj_ccall.c
> @@ -1082,7 +1082,7 @@ static int ccall_set_args(lua_State *L, CTState
> *cts, CType *ct,
> if (CCALL_ALIGN_STACKARG) { /* Align argument on stack. */
> MSize align = (1u << ctype_align(ccall_struct_align(cts, d))) - 1;
> #if LJ_TARGET_ARM64 && LJ_TARGET_OSX
> - isva = ctype_isstruct(d->info);
> + isva |= ctype_isstruct(d->info);
> #endif
> if (rp || (CCALL_PACK_STACKARG && isva && align < CTSIZE_PTR-1))
> align = CTSIZE_PTR-1;
> diff --git
> a/test/tarantool-tests/lj-1455-macos-arm64-vararg-regression.test.lua
> b/test/tarantool-tests/lj-1455-macos-arm64-vararg-regression.test.lua
> new file mode 100644
> index 00000000..a6670cce
> --- /dev/null
> +++ b/test/tarantool-tests/lj-1455-macos-arm64-vararg-regression.test.lua
> @@ -0,0 +1,112 @@
> +local ffi = require('ffi')
> +local tap = require('tap')
> +
> +-- The test file to test various FFI C vararg calls.
> +-- luacheck: push no max_comment_line_length
> +-- Originated from: https://github.com/neovim/neovim/blob/a5aa62e37b82214a1d4d1e0a54d193b155fb340c/test/unit/strings_spec.lua
>
> +-- luacheck: pop
> +-- See also: https://github.com/LuaJIT/LuaJIT/issues/1455.
> +
> +local test = tap.test('lj-1455-macos-arm64-vararg-regression')
> +
> +test:plan(45)
> +
> +ffi.cdef('int sprintf(char *str, const char *format, ...);')
> +
> +local buf = ffi.new('char[64]')
> +
> +local function t(expected, fmt, ...)
> + local args = {...}
> + local ctx = string.format('sprintf(buf, "%s"', fmt)
> + for _, x in ipairs(args) do
> + ctx = ctx .. ', ' .. tostring(x)
> + end
> + ctx = ctx .. string.format(') = %s', expected)
> +
> + test:test(ctx, function(subtest, ...)
> + subtest:plan(2)
> + subtest:is(ffi.C.sprintf(buf, fmt, ...), #expected,
> + ctx .. ' - return status')
> + subtest:is(ffi.string(buf), expected, ctx .. ' - result string')
> + end, ...)
> +end
> +
> +local function i(n)
> + return ffi.cast('int', n)
> +end
> +
> +local function l(n)
> + return ffi.cast('long', n)
> +end
> +
> +local function ll(n)
> + return ffi.cast('long long', n)
> +end
> +
> +local function z(n)
> + return ffi.cast('ptrdiff_t', n)
> +end
> +
> +local function u(n)
> + return ffi.cast('unsigned', n)
> +end
> +
> +local function ul(n)
> + return ffi.cast('unsigned long', n)
> +end
> +
> +local function ull(n)
> + return ffi.cast('unsigned long long', n)
> +end
> +
> +local function uz(n)
> + return ffi.cast('size_t', n)
> +end
> +
> +t('1234567', '%d', i(1234567))
> +t('1234567', '%ld', l(1234567))
> +t(' 1234567', '%9ld', l(1234567))
> +t('1234567 ', '%-9ld', l(1234567))
> +t('deadbeef', '%x', u(0xdeadbeef))
> +t('one two', '%s %s', 'one', 'two')
> +t('1.234000', '%f', 1.234)
> +t('1.234000e+00', '%e', 1.234)
> +t('inf', '%f', 1.0 / 0.0)
> +t('-inf', '%f', -1.0 / 0.0)
> +t('-0.000000', '%f', tonumber('-0.0'))
> +t('%%%', '%%%%%%')
> +t('0x87654321', '%p', ffi.cast('char *', 0x87654321))
> +t('0x0087654321', '%012p', ffi.cast('char *', 0x87654321))
> +t('1234567 ', '%1$*2$ld', l(1234567), i(-9))
> +t('1234567 ', '%1$*2$.*3$ld', l(1234567), i(-9), i(5))
> +t('1234567 ', '%1$*3$.*2$ld', l(1234567), i(5), i(-9))
> +t('1234567 ', '%3$*1$.*2$ld', i(-9), i(5), l(1234567))
> +t('1234567', '%1$ld', l(1234567))
> +t(' 1234567', '%1$*2$ld', l(1234567), i(9))
> +t('9 12345 7654321', '%2$ld %1$d %3$lu', i(12345), l(9), ul(7654321))
> +t('9 1234567 7654321', '%2$d %1$ld %3$lu', l(1234567), i(9), ul(7654321))
>
> +t('9 1234567 7654321', '%2$d %1$lld %3$lu', ll(1234567), i(9),
> ul(7654321))
> +t('9 12345 7654321', '%2$ld %1$u %3$lu', u(12345), l(9), ul(7654321))
> +t('9 1234567 7654321', '%2$d %1$lu %3$lu', ul(1234567), i(9),
> ul(7654321))
> +t('9 1234567 7654321', '%2$d %1$llu %3$lu', ull(1234567), i(9),
> ul(7654321))
> +t('9 deadbeef 7654321', '%2$d %1$x %3$lu', u(0xdeadbeef), i(9),
> ul(7654321))
> +t('9 c 7654321', '%2$ld %1$c %3$lu', i(('c'):byte()), l(9), ul(7654321))
> +t('9 hi 7654321', '%2$ld %1$s %3$lu', 'hi', l(9), ul(7654321))
> +t('9 0.000000e+00 7654321', '%2$ld %1$e %3$lu', 0.0, l(9), ul(7654321))
> +t('two one two', '%2$s %1$s %2$s', 'one', 'two', 'three')
> +t('three one two', '%3$s %1$s %2$s', 'one', 'two', 'three')
> +t('1234567', '%1$d', i(1234567))
> +t('deadbeef', '%1$x', u(0xdeadbeef))
> +t('one two', '%1$s %2$s', 'one', 'two')
> +t('two one', '%2$s %1$s', 'one', 'two')
> +t('1.234000', '%1$f', 1.234)
> +t('1.234000e+00', '%1$e', 1.234)
> +t('inf', '%1$f', 1.0 / 0.0)
> +t('-inf', '%1$f', -1.0 / 0.0)
> +t('-0.000000', '%1$f', tonumber('-0.0'))
> +t('-1234567 -7654321', '%zd %zd', z(-1234567), z(-7654321))
> +t('-7654321 -1234567', '%2$zd %1$zd', z(-1234567), z(-7654321))
> +t('1234567 7654321', '%zu %zu', uz(1234567), uz(7654321))
> +t('7654321 1234567', '%2$zu %1$zu', uz(1234567), uz(7654321))
> +
> +test:done(true)
> --
> 2.54.0
>
[-- Attachment #2: Type: text/html, Size: 7925 bytes --]
next prev parent reply other threads:[~2026-06-30 14:47 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-29 13:19 Sergey Kaplun via Tarantool-patches
2026-06-30 14:46 ` Evgeniy Temirgaleev via Tarantool-patches [this message]
2026-07-01 15:48 ` Sergey Bronnikov via Tarantool-patches
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1782830799.474292866@f378.i.mail.ru \
--to=tarantool-patches@dev.tarantool.org \
--cc=e.temirgaleev@tarantool.org \
--cc=skaplun@tarantool.org \
--subject='Re: [Tarantool-patches] [PATCH luajit] FFI/MacOS: Fix calling convention for on-stack varargs.' \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox