From: Sergey Ostanevich via Tarantool-patches <tarantool-patches@dev.tarantool.org> To: Igor Munkin <imun@tarantool.org> Cc: tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH luajit v1] tools: fix luajit-gdb stack dump Date: Thu, 22 Jul 2021 13:32:23 +0300 [thread overview] Message-ID: <3E1848F9-799D-4A38-9E7C-CD218ADFDF54@tarantool.org> (raw) In-Reply-To: <20210721085411.GJ11494@tarantool.org> Hi! Thanks for the patch! I would rather remove all mentions of ‘bottom loop’ as we named it at Intel. I'm not a big pro in python, still the ‘for’ loop works fine for me: If we can unwind to the next frame - dump it (and further) >>> s=[1] >>> for i in s: ... if (i==1): ... s.append(2) ... continue ... print(i) ... 2 If it is the only one (dummy): >>> s=[3] >>> for i in s: ... if (i==1): ... s.append(2) ... continue ... print(i) ... 3 Does it make sense in this case? Anyways, it’s too much of nitpicking - LGTM in its current version. Regards, Sergos > On 21 Jul 2021, at 11:54, Igor Munkin <imun@tarantool.org> wrote: > > Sergey, > > Thanks for the fixes! Everything is perfect now, I'll ask Sergos to > proceed with the review. > > On 21.07.21, Sergey Kaplun wrote: >> Igor, >> >> On 20.07.21, Igor Munkin wrote: >>> Sergey, >>> >>> Thanks for the patch! Nice catch, actually! I wonder, how much impact >>> this bug made while digging the crash reports. Anyway, LGTM, but >>> consider several nits below. >> >> Actually, not too much. It is the unique situation, when we have naked >> guest Lua stack without any function call. > > Hm, I guess this bug affects all the cases when dummy frame was used, > but you know it better anyway :) > >> >> Branch is force-pushed with the changes below. >> >>> >>> On 18.06.21, Sergey Kaplun wrote: > > <snipped> > >> >> Thanks for your comments! >> The new commit message version: >> >> | Dummy frame is the "initial" coroutine state, when the framelink slot (i.e. >> | L->base - (1 + LJ_FR2)) is the bottom slot of the guest stack >> | (i.e. L->stack). Since coroutine stack unwinding is implemented via >> | precondition loop, lj-stack doesn't dump the slots for the dummy frame, since >> | the framelink points to the stack bottom. >> | >> | The output looks like the following: >> | >> | | 0x7fb512ac40:0x7fb512ac70 [ ] 7 slots: Red zone >> | | 0x7fb512ac38 [ M] >> | | 0x7fb512ab28:0x7fb512ac30 [ ] 34 slots: Free stack slots >> | | 0x7fb512ab20 [ T ] >> | | 0x7fb512ab08:0x7fb512ab10 [S ] FRAME: dummy L >> | >> | Python doesn't provide post-condition (do-while) syntax construction, >> | that fits better for this case, so the unwinding of the topmost frame is just manually >> | unrolled. >> | >> | As a result of the patch the output looks like the following: >> | >> | | 0x7fb512ac40:0x7fb512ac70 [ ] 7 slots: Red zone >> | | 0x7fb512ac38 [ M] >> | | 0x7fb512ab28:0x7fb512ac30 [ ] 34 slots: Free stack slots >> | | 0x7fb512ab20 [ T ] >> | | 0x7fb512ab18 [ ] VALUE: string 0 "/tmp/net_box.lua:6: err in ser" @ 0x7fb512ade8 >> | | 0x7fb512ab10 [ B ] VALUE: table @ 0x7fb512ac80 (asize: 0, hmask: 0x0) >> | | 0x7fb512ab00:0x7fb512ab08 [S ] FRAME: dummy L > > Side note: I guess this is a previous version, since everything is fine > on the branch. BTW, wording is neat! > >> > > <snipped> > >>>> diff --git a/src/luajit-gdb.py b/src/luajit-gdb.py >>>> index f1fd6230..a3dad585 100644 >>>> --- a/src/luajit-gdb.py >>>> +++ b/src/luajit-gdb.py >>>> @@ -424,16 +424,26 @@ def dump_stack(L, base=None, top=None): > > <snipped> > >> >>> >>>> while framelink > mref('TValue *', L['stack']): >>>> - while slot > framelink + LJ_FR2: >>>> - dump += dump_stack_slot(L, slot, base, top) >>>> - slot -= 1 >>>> + assert slot == framelink + LJ_FR2, "Invalid slot during frame unwind" >>> >>> Please, use parentheses for assert call. >> >> It's done by design to avoid always true assertions [1]: >> >> | $ python3 >> | Python 3.7.10 (default, Mar 18 2021, 04:43:06) >> | [GCC 9.3.0] on linux >> | Type "help", "copyright", "credits" or "license" for more information. >> | >>> assert(1!=1, "stupid assert") >> | <stdin>:1: SyntaxWarning: assertion is always true, perhaps remove parentheses? > > PEBKAC, didn't expect such a surprise here. I looked on the <assert> > usage below, but those parentheses are just excess. Everything is fine > here then, and thanks for clarification! > >> > > <snipped> > >> >> [1]: https://docs.python.org/3/reference/simple_stmts.html#grammar-token-assert-stmt >> >> -- >> Best regards, >> Sergey Kaplun > > -- > Best regards, > IM
next prev parent reply other threads:[~2021-07-22 10:32 UTC|newest] Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-06-18 6:22 Sergey Kaplun via Tarantool-patches 2021-07-05 7:03 ` Sergey Kaplun via Tarantool-patches 2021-07-20 12:29 ` Igor Munkin via Tarantool-patches 2021-07-21 8:49 ` Sergey Kaplun via Tarantool-patches 2021-07-21 8:54 ` Igor Munkin via Tarantool-patches 2021-07-22 10:32 ` Sergey Ostanevich via Tarantool-patches [this message] 2021-07-22 11:48 ` Igor Munkin via Tarantool-patches 2021-07-22 12:51 ` Sergey Ostanevich via Tarantool-patches 2021-07-21 9:02 ` Sergey Kaplun via Tarantool-patches 2021-07-22 13:34 ` 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=3E1848F9-799D-4A38-9E7C-CD218ADFDF54@tarantool.org \ --to=tarantool-patches@dev.tarantool.org \ --cc=imun@tarantool.org \ --cc=sergos@tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH luajit v1] tools: fix luajit-gdb stack dump' \ /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