Tarantool development patches archive
 help / color / mirror / Atom feed
From: Sergey Bronnikov via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: Sergey Kaplun <skaplun@tarantool.org>,
	Maxim Kokryashkin <m.kokryashkin@tarantool.org>
Cc: tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH luajit v1 3/5] test: allow `jit.parse` to return aborted traces
Date: Thu, 6 Jun 2024 14:01:37 +0300	[thread overview]
Message-ID: <42b64daf-a9f6-4267-9d22-4c8d7b88fdca@tarantool.org> (raw)
In-Reply-To: <7c0a7be30070d085d6ddd3b56c4338686e1c2baa.1713773432.git.skaplun@tarantool.org>

[-- Attachment #1: Type: text/plain, Size: 4199 bytes --]

Sergey,


thanks for the patch! Please see my comments below.

On 22.04.2024 11:49, Sergey Kaplun wrote:
> Now information about the abort of the trace is saved in the
> `abort_reason` field of the corresponding structure. The
> `jit.parse.finish()` returns now the second table containing aborted
> traces. Each table key is a trace number containing an array of
> potentially traces with this number, which was aborted.
>
> Needed for tarantool/tarantool#9924
> ---
>   .../unit-jit-parse-abort.test.lua             | 38 +++++++++++++++++++
>   test/tarantool-tests/utils/jit/parse.lua      | 22 ++++++++---
>   2 files changed, 55 insertions(+), 5 deletions(-)
>   create mode 100644 test/tarantool-tests/unit-jit-parse-abort.test.lua
>
> diff --git a/test/tarantool-tests/unit-jit-parse-abort.test.lua b/test/tarantool-tests/unit-jit-parse-abort.test.lua
> new file mode 100644
> index 00000000..91af5a56
> --- /dev/null
> +++ b/test/tarantool-tests/unit-jit-parse-abort.test.lua
> @@ -0,0 +1,38 @@
> +local tap = require('tap')
> +local test = tap.test('unit-jit-parse'):skipcond({

Usually a name passed to `tap.test` matches to test file name,

but here it is not matched.

> +  ['Test requires JIT enabled'] = not jit.status(),
> +  ['Disabled on *BSD due to #4819'] = jit.os == 'BSD',
> +})
> +
> +local jparse = require('utils').jit.parse
> +
> +-- XXX: Avoid other traces compilation due to hotcount collisions
> +-- for predictable results.
> +jit.off()
> +jit.flush()
> +
> +test:plan(1)
> +
> +jit.on()
> +-- We only need the abort reason in the test.
> +jparse.start('t')

I would add a comment with explanation what does 't' flag mean.

Feel free to ignore.

> +
> +-- XXX: A trace always has at least 3 IR constants: for `nil`,
> +-- `false`, and `true`. Always fails to record with the set
> +-- `maxirconst` limit.
> +jit.opt.start('hotloop=1', 'maxirconst=1')
> +
> +for _ = 1, 3 do end
> +
> +local _, aborted_traces = jparse.finish()
> +
> +jit.off()
> +
> +assert(aborted_traces and aborted_traces[1], 'aborted trace is persisted')
> +
> +-- We tried to compile only one trace.
> +local reason = aborted_traces[1][1].abort_reason
> +
> +test:like(reason, 'trace too long', 'abort reason is correct')
> +
> +test:done(true)
> diff --git a/test/tarantool-tests/utils/jit/parse.lua b/test/tarantool-tests/utils/jit/parse.lua
> index bcef5b35..0ce7f7c8 100644
> --- a/test/tarantool-tests/utils/jit/parse.lua
> +++ b/test/tarantool-tests/utils/jit/parse.lua
> @@ -22,6 +22,7 @@ local function trace_new(n)
>       parent = nil,
>       parent_exitno = nil,
>       is_stitched = false,
> +    abort_reason = nil,
>       start_loc = nil,
>       bc = {},
>       ir = {},
> @@ -87,9 +88,17 @@ local header_handlers = {
>       ctx.parsing_trace = nil
>       ctx.parsing = nil
>     end,
> -  abort = function(ctx, trace_num)
> +  abort = function(ctx, trace_num, line)
>       local traces = ctx.traces
>       assert(ctx.parsing_trace == trace_num)
> +
> +    local aborted_traces = ctx.aborted_traces
> +    if not aborted_traces[trace_num] then
> +      aborted_traces[trace_num] = {}
> +    end
> +    traces[trace_num].abort_reason = line:match('-- (.+)$')
> +    table.insert(aborted_traces[trace_num], traces[trace_num])
> +
>       ctx.parsing_trace = nil
>       ctx.parsing = nil
>       traces[trace_num] = nil
> @@ -137,11 +146,14 @@ end
>   local JDUMP_FILE
>   
>   local function parse_jit_dump()
> -  local ctx = {traces = {}}
> +  local ctx = {
> +    aborted_traces = {},
> +    traces = {},
> +  }
>     for line in io.lines(JDUMP_FILE) do
>       parse_line(ctx, line)
>     end
> -  return ctx.traces
> +  return ctx.traces, ctx.aborted_traces
>   end
>   
>   -- Start `jit.dump()` utility with the given flags, saving the
> @@ -167,10 +179,10 @@ M.finish = function()
>     -- Enable traces compilation for `jit.dump` back.
>     jit.on(jdump.on, true)
>     jit.on(jdump.off, true)
> -  local traces = parse_jit_dump()
> +  local traces, aborted_traces = parse_jit_dump()
>     os.remove(JDUMP_FILE)
>     JDUMP_FILE = nil
> -  return traces
> +  return traces, aborted_traces
>   end
>   
>   -- Turn off compilation for the module to avoid side effects.

[-- Attachment #2: Type: text/html, Size: 4869 bytes --]

  parent reply	other threads:[~2024-06-06 11:01 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-22  8:49 [Tarantool-patches] [PATCH luajit v1 0/5] Handle all errors during stitching Sergey Kaplun via Tarantool-patches
2024-04-22  8:49 ` [Tarantool-patches] [PATCH luajit v1 1/5] build: introduce option LUAJIT_ENABLE_TABLE_BUMP Sergey Kaplun via Tarantool-patches
2024-05-05 12:39   ` Maxim Kokryashkin via Tarantool-patches
2024-05-13 11:47     ` Sergey Kaplun via Tarantool-patches
2024-06-06 10:53   ` Sergey Bronnikov via Tarantool-patches
2024-06-13  9:51     ` Sergey Kaplun via Tarantool-patches
2024-06-13 14:37       ` Sergey Bronnikov via Tarantool-patches
2024-04-22  8:49 ` [Tarantool-patches] [PATCH luajit v1 2/5] ci: add tablebump flavor for exotic builds Sergey Kaplun via Tarantool-patches
2024-05-05 12:40   ` Maxim Kokryashkin via Tarantool-patches
2024-05-13 11:49     ` Sergey Kaplun via Tarantool-patches
2024-06-06 10:56   ` Sergey Bronnikov via Tarantool-patches
2024-04-22  8:49 ` [Tarantool-patches] [PATCH luajit v1 3/5] test: allow `jit.parse` to return aborted traces Sergey Kaplun via Tarantool-patches
2024-05-05 12:55   ` Maxim Kokryashkin via Tarantool-patches
2024-05-13 11:53     ` Sergey Kaplun via Tarantool-patches
2024-06-06 11:01   ` Sergey Bronnikov via Tarantool-patches [this message]
2024-06-13 10:00     ` Sergey Kaplun via Tarantool-patches
2024-06-13 14:38       ` Sergey Bronnikov via Tarantool-patches
2024-04-22  8:49 ` [Tarantool-patches] [PATCH luajit v1 4/5] Handle all types of errors during trace stitching Sergey Kaplun via Tarantool-patches
2024-05-06  8:25   ` Maxim Kokryashkin via Tarantool-patches
2024-06-06 13:03   ` Sergey Bronnikov via Tarantool-patches
2024-06-13 10:25     ` Sergey Kaplun via Tarantool-patches
2024-06-13 14:52       ` Sergey Bronnikov via Tarantool-patches
2024-04-22  8:49 ` [Tarantool-patches] [PATCH luajit v1 5/5] Use generic trace error for OOM " Sergey Kaplun via Tarantool-patches
2024-05-06  8:27   ` Maxim Kokryashkin via Tarantool-patches
2024-06-06 14:06   ` Sergey Bronnikov via Tarantool-patches
2024-06-13 10:31     ` Sergey Kaplun via Tarantool-patches
2024-06-13 14:58       ` Sergey Bronnikov via Tarantool-patches
2024-06-16 19:13         ` Sergey Kaplun 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=42b64daf-a9f6-4267-9d22-4c8d7b88fdca@tarantool.org \
    --to=tarantool-patches@dev.tarantool.org \
    --cc=m.kokryashkin@tarantool.org \
    --cc=sergeyb@tarantool.org \
    --cc=skaplun@tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH luajit v1 3/5] test: allow `jit.parse` to return aborted traces' \
    /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