From: Maxim Kokryashkin via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: "Sergey Kaplun" <skaplun@tarantool.org>
Cc: "Sergey Bronnikov" <estetus@gmail.com>,
tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH luajit v1] Fix BC_UCLO insertion for returns.
Date: Wed, 07 Jun 2023 14:35:25 +0300 [thread overview]
Message-ID: <1686137725.949044142@f485.i.mail.ru> (raw)
In-Reply-To: <ZH8ryLJJe/Ar2x7C@root>
[-- Attachment #1: Type: text/plain, Size: 4108 bytes --]
Hi, Sergey and Sergey!
>
>>Hi, Sergey!
>>Thanks for the patch!
>>Please, consider my comments below.
>>
>>On 30.05.23, Sergey Bronnikov wrote:
>>> From: Sergey Bronnikov < sergeyb@tarantool.org >
>>>
>>> Contributed by XmiliaH.
>>>
>>> (cherry-picked from commit 93a65d3cc263aef2d2feb3d7ff2206aca3bee17e)
>>>
>>> After emitting bytecode instruction BC_FNEW fixup is not required,
>Typo: s/bytecode/the bytecode
>>> because FuncState will set a flag PROTO_CHILD that will trigger emitting
>>> a pair of instructions BC_UCLO and BC_RET (see <src/lj_parse.c:2355>)
>>> and BC_RET will close all upvalues from base equal to 0.
>>
>>This part describes why replacing UCLO with FNEW is good enough and
>>better than just deleting
>>| case BC_UCLO: return;
>>But the original problem is that some of BC_RET are not fixup-ed, due to
>>early return, if UCLO is obtained before, those leads to VM
>>inconsistency after return from the function. Please, mention this too.
>Agree here, it is hard to get what the patch is about from that description,
>without diving into the changes.
>>>
>>> Sergey Bronnikov:
>>> * added the description and the test for the problem
>>>
>>> Signed-off-by: Sergey Bronnikov < sergeyb@tarantool.org >
>>> Co-authored-by: Sergey Kaplun < skaplun@tarantool.org >
>>> ---
>>> Branch: https://github.com/tarantool/luajit/tree/ligurio/gh-819-fix-missing-uclo
>>> PR: https://github.com/tarantool/tarantool/pull/8689
>>>
>>> src/lj_parse.c | 2 +-
>>> .../lj-819-fix-missing-uclo.test.lua | 27 +++++++++++++++++++
>>> 2 files changed, 28 insertions(+), 1 deletion(-)
>>> create mode 100644 test/tarantool-tests/lj-819-fix-missing-uclo.test.lua
>>>
>>> diff --git a/src/lj_parse.c b/src/lj_parse.c
>>> index af0dc53f..343fa797 100644
>>> --- a/src/lj_parse.c
>>> +++ b/src/lj_parse.c
>>
>><snipped>
>>
>>> diff --git a/test/tarantool-tests/lj-819-fix-missing-uclo.test.lua b/test/tarantool-tests/lj-819-fix-missing-uclo.test.lua
>>> new file mode 100644
>>> index 00000000..b3f1f78a
>>> --- /dev/null
>>> +++ b/test/tarantool-tests/lj-819-fix-missing-uclo.test.lua
>>> @@ -0,0 +1,27 @@
>>> +local tap = require('tap')
>>> +local test = tap.test('lj-819-fix-missing-uclo')
>>> +
>>> +test:plan(1)
>>> +
>>> +local function missing_uclo()
>>> + while true do -- luacheck: ignore
>>> + if false then
>>> + break
>>
>>Please, comment why do we need this always false branch for reproducer
>>(the aforementioned BC_UCLO).
>>
>>Also, examples of bytecode listings for this function before and after
>>the patch are desirable.
>>
>>> + end
>>> + local f
>>> + while true do
>>> + if f then
>>> + return f
>>
>>Please, comment, that exactly here we got not fixupped RET before the
>>patch.
>>
>>> + end
>>> + f = function()
>>> + return f
>>> + end
>>> + end
>>> + end
>>> +end
>>> +
>>> +local f = missing_uclo()
>>> +local res = f()
>>> +test:ok(type(res) == 'function', 'type of returned value is correct')
>>
>>Minor: the comment why we don't get here a function, when upvalue isn't
>>closed is desirable.
>>
>>> +
>>> +os.exit(test:check() and 0 or 1)
>>
>>Also, before the patch I got the following assertion in JIT:
>>
>>| LUA_PATH="src/?.lua;;" src/luajit -Ohotloop=1 -e '
>>|
>>| local function missing_uclo()
>>| while true do -- luacheck: ignore
>>| local f
>>| if false then break end
>>| while true do
>>| if f then
>>| return f
>>| end
>>| f = function()
>>| return f
>>| end
>>| end
>>| end
>>| end
>>| f = missing_uclo()
>>| print(f())
>>| f = missing_uclo()
>>| print(f())
>>| '
>>| 3.1002202036551
>>| luajit: /home/burii/reviews/luajit/lj-819-missing-uclo/src/lj_record.c:135: rec_check_slots: Assertion `((((((tr))>>24) & IRT_TYPE) - (TRef)(IRT_NUM) <= (TRef)
>>| (IRT_INT-IRT_NUM)))' failed.
>>| Aborted
>>
>>I don't sure that we should test this particular failure too, since the
>>origin of the problem is the incorrect emitted bytecode.
>>
>>Thoughts?
>We should not, because it is most likely caused by the issue
>that was fixed in the LuaJIT/LuaJIT@5c46f477.
>>
>>> --
>>> 2.34.1
>>>
>>
>>--
>>Best regards,
>>Sergey Kaplun
>--
>Best regards,
>Maxim Kokryashkin
>
[-- Attachment #2: Type: text/html, Size: 5800 bytes --]
next prev parent reply other threads:[~2023-06-07 11:35 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-05-30 16:56 Sergey Bronnikov via Tarantool-patches
2023-06-06 12:51 ` Sergey Kaplun via Tarantool-patches
2023-06-07 11:35 ` Maxim Kokryashkin via Tarantool-patches [this message]
2023-07-06 9:43 ` Sergey Bronnikov via Tarantool-patches
2023-07-06 11:31 ` Maxim Kokryashkin via Tarantool-patches
2023-07-06 13:45 ` Sergey Bronnikov via Tarantool-patches
2023-07-06 21:12 ` Maxim Kokryashkin via Tarantool-patches
2023-07-06 9:40 ` Sergey Bronnikov via Tarantool-patches
2023-07-09 13:15 ` Sergey Kaplun via Tarantool-patches
2023-07-10 14:53 ` Sergey Bronnikov via Tarantool-patches
2023-07-13 7:57 ` Sergey Kaplun via Tarantool-patches
2023-07-13 9:55 ` Sergey Bronnikov via Tarantool-patches
2023-07-13 10:25 ` Sergey Kaplun via Tarantool-patches
2023-07-20 18:37 ` Igor Munkin 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=1686137725.949044142@f485.i.mail.ru \
--to=tarantool-patches@dev.tarantool.org \
--cc=estetus@gmail.com \
--cc=m.kokryashkin@tarantool.org \
--cc=skaplun@tarantool.org \
--subject='Re: [Tarantool-patches] [PATCH luajit v1] Fix BC_UCLO insertion for returns.' \
/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