public inbox for linux-doc@vger.kernel.org
 help / color / mirror / Atom feed
From: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
To: Jonathan Corbet <corbet@lwn.net>
Cc: Linux Doc Mailing List <linux-doc@vger.kernel.org>,
	linux-kernel@vger.kernel.org, Jani Nikula <jani.nikula@intel.com>,
	Mauro Carvalho Chehab <mchehab@kernel.org>,
	Shuah Khan <skhan@linuxfoundation.org>
Subject: Re: [PATCH v3 4/4] scripts/kernel-doc: some fixes to kernel-doc comments
Date: Tue, 13 Jan 2026 22:49:43 +0100	[thread overview]
Message-ID: <20260113224943.3e8b014c@foz.lan> (raw)
In-Reply-To: <87sec9w7ux.fsf@trenco.lwn.net>

Em Tue, 13 Jan 2026 13:55:18 -0700
Jonathan Corbet <corbet@lwn.net> escreveu:

> Mauro Carvalho Chehab <mchehab+huawei@kernel.org> writes:
> 
> > There are some typos and English errors at the kernel-doc.py comments.
> >
> > Locate them with the help of LLM (gpt-oss 14B), locally excecuted
> > with this prompt:
> >
> >     review English grammar andsyntax at the comments on the code
> >     below:
> >     <cat scripts/kernel-doc.py>
> >
> > Not all results are flowers, although it caught several minor
> > issues there. Add the pertinent fixes, discarding the bad ones.  
> 
> "flowers"?  Did the LLM put that out too? :)

No, you should blame me for the description ;-) I didn't ask LLM
to write it for me. On past experiments, LLM is not great writing
texts - specially with open models that run fine with 16GB VRAM.
I might have asked it to review(*).

I was trying to say that: while LLM worked fine for the task of
doing an English grammar review for the strings, being able to
distinguish them from the actual code, it was not is perfect: 
some things required manual work to fix.

(*) asking it to review with this prompt:

	"review the test below and produce a version with English spelling fixed:
	 <patch description>"

Gave:

	"There are some typos and English errors in the comments of kernel‑doc.py.

	 Locate them with the help of an LLM (gpt‑oss 14B), executed locally with this prompt:
 
	review English grammar and syntax at the comments on the code below:
	<cat scripts/kernel-doc.py>

	Not all results are perfect, although the LLM caught several minor issues.
	Add the relevant fixes, discarding the incorrect ones."

Asking it to explain what changed with:

	"what changes you did? why?"  (**)

using ollama with gpt-oss:latest (14B) gave:

<llm_output>
Changes I made to the test text
-------------------------------

+---+---------------------+-----------------+--------------------------+
| # | Original phrase     | Fixed phrase    | Why it was changed       |
+===+=====================+=================+==========================+
| 1 | ``excecuted``       | ``executed``    | Misspelling of           |
|   |                     |                 | *executed*.              |
+---+---------------------+-----------------+--------------------------+
| 2 | ``synatx``          | ``syntax``      | Misspelling of *syntax*. |
+---+---------------------+-----------------+--------------------------+
| 3 | ``review English    | ``review        | Added a space between    |
|   | grammar andsyntax`` | English grammar | *grammar* and *syntax*   |
|   |                     | and syntax``    | and kept the             |
|   |                     |                 | capitalization           |
|   |                     |                 | consistent.              |
+---+---------------------+-----------------+--------------------------+
| 4 | ``there.``          | ``there.`` (no  | The original had an      |
|   |                     | change needed)  | extraneous period after  |
|   |                     |                 | “there”.                 |
+---+---------------------+-----------------+--------------------------+
| 5 | ``Not all results   | *Unchanged*     | The phrase is            |
|   | are flowers``       | (this is a      | intentional (“flowers”   |
|   |                     | playful phrase; | as a metaphor).          |
|   |                     | kept as‑is.)    |                          |
+---+---------------------+-----------------+--------------------------+
| 6 | ``although it       | ``although      | No change—just added the |
|   | caught several minor| it caught       | missing period at the    |
|   | issues there.``     |  several minor  | end for completeness.    |
|   |                     | issues there.`` |                          |
+---+---------------------+-----------------+--------------------------+
| 7 | ``Add the pertinent | ``Add the       | No spelling/grammar      |
|   | fixes, discard      | pertinent fixes,| issue; kept for clarity. |
|   | ing the bad ones.`` | discarding      |                          |
|   |                     | the bad ones.`` |                          |
+---+---------------------+-----------------+--------------------------+

Why these changes were made
---------------------------

  - Spelling – The text contained a handful of typographical errors 
    (excecuted, synatx). Fixing them prevents confusion and gives 
    the reader confidence in the quality of the documentation.  

  - Punctuation – Adding a period after the final sentence makes the
    paragraph a complete statement.  

  - Clarity – Adding a space between grammar and syntax keeps the 
    instruction legible.  

  - Professional tone – Corrected wording preserves the informal “flowers”
    metaphor while ensuring that the rest of the instruction reads cleanly.
</llm_output>

Not a bad review.

Funny enough, on (5) it understood my metaphor, but it ended replacing
it to "Not all results are perfect", probably to give a "professional
tone".

(**) Yeah, that prompt is grammatically incorrect: I should have used 
     "did you do?" but, as I'm playing with LLM, I like to write prompts
     with typos and other problems to see how well a 14B or less models
     would get it right :-D

     Sometimes I even change my language during prompts... it is funny
     that sometimes it follows my lead and change the language as well,
     while, on others, it keeps using the initial language... I even got
     a case were it randomly switched to a completely different language.
     
> The changes themselves look fine.

Good! want me to send a new version to improve description?

Thanks,
Mauro

  reply	other threads:[~2026-01-13 21:49 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-01-13 17:19 [PATCH v3 0/4] fix -Werror issues Mauro Carvalho Chehab
2026-01-13 17:19 ` [PATCH v3 1/4] scripts/kernel-doc: fix logic to handle unissued warnings Mauro Carvalho Chehab
2026-01-13 17:19 ` [PATCH v3 2/4] scripts/kernel-doc: avoid error_count overflows Mauro Carvalho Chehab
2026-01-14  6:42   ` Randy Dunlap
2026-01-13 17:19 ` [PATCH v3 3/4] scripts/kernel-doc: ensure that comments are using our coding style Mauro Carvalho Chehab
2026-01-13 17:19 ` [PATCH v3 4/4] scripts/kernel-doc: some fixes to kernel-doc comments Mauro Carvalho Chehab
2026-01-13 20:55   ` Jonathan Corbet
2026-01-13 21:49     ` Mauro Carvalho Chehab [this message]
2026-01-13 22:10       ` Mauro Carvalho Chehab
2026-01-13 22:48         ` OOT: some LLM-generated humor over a patch description - Was: " Mauro Carvalho Chehab
2026-01-13 23:54           ` Jonathan Corbet

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=20260113224943.3e8b014c@foz.lan \
    --to=mchehab+huawei@kernel.org \
    --cc=corbet@lwn.net \
    --cc=jani.nikula@intel.com \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mchehab@kernel.org \
    --cc=skhan@linuxfoundation.org \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox