From: "Antonin Godard" <antonin.godard@bootlin.com>
To: "Quentin Schulz" <quentin.schulz@cherry.de>,
<docs@lists.yoctoproject.org>
Cc: "Thomas Petazzoni" <thomas.petazzoni@bootlin.com>
Subject: Re: [docs] [PATCH 6/6] ref-manual/variables: improve the UNPACKDIR documentation
Date: Mon, 24 Feb 2025 11:53:37 +0100 [thread overview]
Message-ID: <D80MBXN4BSZF.2OK87WROBFBJP@bootlin.com> (raw)
In-Reply-To: <f4db28cd-1701-49ba-8c90-938128fe61df@cherry.de>
Hi Quentin,
On Thu Feb 20, 2025 at 5:41 PM CET, Quentin Schulz wrote:
> Hi Antonin,
>
> On 2/18/25 10:12 AM, Antonin Godard via lists.yoctoproject.org wrote:
>> It was clear why UNPACKDIR was introduced at first, and what is the
>> recommended way of setting S and UNPACKDIR for a clean separation of the
>> source code and WORKDIR. Add documentation for this in the reference
>> manual.
>>
>> Signed-off-by: Antonin Godard <antonin.godard@bootlin.com>
>> ---
>> documentation/ref-manual/variables.rst | 39 +++++++++++++++++++++++++++++++++-
>> 1 file changed, 38 insertions(+), 1 deletion(-)
>>
>> diff --git a/documentation/ref-manual/variables.rst b/documentation/ref-manual/variables.rst
>> index adbef69d8f39d33be87c5db6688a807156540410..d970fc21f69fe35830a9d6b5a28da7cb257c0709 100644
>> --- a/documentation/ref-manual/variables.rst
>> +++ b/documentation/ref-manual/variables.rst
>> @@ -10035,9 +10035,46 @@ system and gives an overview of their function and contents.
>>
>> :term:`UNPACKDIR`
>> This variable, used by the :ref:`ref-classes-base` class,
>> - specifies where fetches sources should be unpacked by the
>> + specifies where fetched sources should be unpacked by the
>> :ref:`ref-tasks-unpack` task.
>>
>> + Although the default value of this directory is
>> + ``${WORKDIR}/sources-unpack``, the recommended way of setting
>> + :term:`UNPACKDIR` and :term:`S` is::
>> +
>> + S = "${WORKDIR}/sources"
>> + UNPACKDIR = "${S}"
>> +
>
> Mmmm I think this is an issue, because we then show a non-recommended
> configuration in the SVG in the previous patches where UNPACKDIR != S.
I'd rather show the default definitions in the figures, because there's no
guarantee that a recipe will define UNPACKDIR and S that way.
It also feels weird to me that a recommended approach is not the default, but it
is probably related to backward compatiblity, that would be my guess.
>> + This allows the source code to be safely unpacked and patched in a
>> + separate directory.
>> +
>
> I think I know where you want to go with this, but the first time I read
> this I didn't understand it. UNPACKDIR = S in the example above, why
> does this say the source code is unpacked and patched (implying two
> steps/stages) in a separate directory then?
I wanted to say that the code is unpacked and patched in the same directory, but
separate from the WORKDIR. I should probably rephrase or remove this sentence to
avoid adding confusion.
> I think this is because we used to put everything from the fetcher into
> WORKDIR and there were unexpected leftovers between builds and UNPACKDIR
> now is a separate directory for the fetcher to put everything it does
> in, instead of WORKDIR?
Yes, I think that's it.
> I'm not sure we need to document how messed up it was in the past rather
> just what this does now?
So I guess we can go with the sentence removal. However I think we'd probably
want to keep the recommended assignment.
>> + .. note::
>> +
>> + In some cases, the :term:`UNPACKDIR` directory only holds the sources
>> + temporarily. When the first directory after :term:`WORKDIR` in the
>> + :term:`S` variable is matched in :term:`UNPACKDIR`, the directory
>> + in :term:`UNPACKDIR` is moved to the :term:`WORKDIR` directory.
>> +
>> + For example, let's represent the :term:`UNPACKDIR` directory
>> + for the ``foobar`` recipe after extraction as follows::
>> +
>> + ${WORKDIR}/sources-unpack/foobar-1.0.0/...
>> +
>> + And the variable :term:`S` for this recipe defined as::
>> +
>> + S = "${WORKDIR}/${BP}"
>> +
>> + Where :term:`BP` expands to ``foobar-1.0.0``.
>> +
>> + Since the first directory of :term:`S` after :term:`WORKDIR`
>> + (``foobar-1.0.0``) matches the directory ``foobar-1.0.0`` under
>> + ``sources-unpack``, the latter is moved to the :term:`WORKDIR`
>> + directory. The final location of the source code will then be
>> + ``${WORKDIR}/foobar-1.0.0``.
>> +
>> + This behavior was added for compatibility for recipes that do not set
>> + the :term:`UNPACKDIR` and :term:`S` variable in the recommended way.
>> +
>
> I've read this 5 times now, it may be too late for my brain today but I
> am not sure I got it :)
>
> Let me try to reword it just to make sure I understood/got it right?
>
> If UNPACKDIR differs from S, but the name of the directory (dirA) where
> sources are unpacked in UNPACKDIR matches the basename of the directory
> used for S, then the content of UNPACKDIR/dirA is moved into S?
That's it. And it is a pain to describe textually. I'd take any recommendations
for improving it.
> If that's right, what really happens? Is UNPACKDIR variable modified to
> point to S once the move is done? When is this happening?
No, UNPACKDIR isn't modified. This move is repeated at the end of the do_unpack
task, each time. See:
https://git.openembedded.org/openembedded-core/tree/meta/classes-global/base.bbclass#n158
I may consider removing that note entirely depending on how useful it is to
document a backwards-compatibility behavior.
Antonin
--
Antonin Godard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
prev parent reply other threads:[~2025-02-24 10:53 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-18 9:11 [PATCH 0/6] Updates around the UNPACKDIR variable Antonin Godard
2025-02-18 9:11 ` [PATCH 1/6] overview-manual: convert source-fetching.png to svg and fix UNPACKDIR Antonin Godard
2025-02-20 16:03 ` [docs] " Quentin Schulz
2025-02-18 9:11 ` [PATCH 2/6] overview-manual: convert patching.png to svg Antonin Godard
2025-02-20 16:05 ` [docs] " Quentin Schulz
2025-02-18 9:11 ` [PATCH 3/6] overview-manual: convert configuration-compile-autoreconf.png " Antonin Godard
2025-02-20 16:13 ` [docs] " Quentin Schulz
2025-02-24 9:19 ` Antonin Godard
2025-02-24 10:26 ` Quentin Schulz
2025-02-24 10:37 ` Antonin Godard
2025-02-18 9:11 ` [PATCH 4/6] overview-manual: convert analysis-for-package-splitting.png " Antonin Godard
2025-02-20 16:18 ` [docs] " Quentin Schulz
2025-02-24 10:44 ` Antonin Godard
2025-02-24 10:48 ` Quentin Schulz
2025-02-18 9:12 ` [PATCH 5/6] overview-manual: add UNPACKDIR in the directory description Antonin Godard
2025-02-20 16:22 ` [docs] " Quentin Schulz
2025-02-18 9:12 ` [PATCH 6/6] ref-manual/variables: improve the UNPACKDIR documentation Antonin Godard
2025-02-20 16:41 ` [docs] " Quentin Schulz
2025-02-24 10:53 ` Antonin Godard [this message]
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=D80MBXN4BSZF.2OK87WROBFBJP@bootlin.com \
--to=antonin.godard@bootlin.com \
--cc=docs@lists.yoctoproject.org \
--cc=quentin.schulz@cherry.de \
--cc=thomas.petazzoni@bootlin.com \
/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