Openembedded Core Discussions
 help / color / mirror / Atom feed
From: Richard Purdie <richard.purdie@linuxfoundation.org>
To: Peter Suti <peter.suti@streamunlimited.com>
Cc: openembedded-core@lists.openembedded.org
Subject: Re: [OE-core] [PATCH] externalsrc: fix dependency chain issues
Date: Fri, 04 Aug 2023 15:47:31 +0100	[thread overview]
Message-ID: <aeed5cf0689586cedccac7d9be2b36d5f5d6c8ad.camel@linuxfoundation.org> (raw)
In-Reply-To: <CACMGZgazaZR1z4opzxwNNug-97DanqC6qciEzS+noshjQcsTeg@mail.gmail.com>

On Wed, 2023-08-02 at 11:26 +0200, Peter Suti wrote:
> On Mon, Jul 31, 2023 at 2:33 PM Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> > 
> > I did look at the bug and it sounds like if you clean a recipe, builds
> > of other recipes which depend upon it then fail?
> 
> Well not just in this case specifically, it is reproducible when doing
> a clean build.
> But this is the easiest most convenient way to reproduce because most
> of the sstate can be reused.

Do you have a way to reproduce this using recipes in OE-Core?

For example, I tried adding this to the cairo recipe:

EXTERNALSRC = "/xxx/cairo-1.16.0"
inherit externalsrc

and then building cairo, cleaning cairo, building pango (which depends
on cairo). I couldn't see any failure.

> > We need to understand what is happening with both the sstate manifests
> > and the stamp files to fully understand this bug.
> > 
> > do_prepare_recipe_sysroot of this recipe that depends on the
> > externalsrc one should trigger it's do_populate_sysroot. The question
> > is whether the do_clean removed the do_populate_sysroot stamp file for
> > that recipe or not? It sounds like it did remove the sstate manifest.
> 
> I have checked the manifest and stamp creation and as far as I can
> tell that part is correct.
> 
> > The more I look at this, the more I think changing the deltask is just
> > masking the issue and there is some other problem at play here which we
> > need to address directly.
> 
> I agree, I have dug a bit further and the issue seems to come from runqueue.py.
> It seems that the current implementation depends heavily on the
> presence of _setscene tasks to build the dependency graph. Altering
> this behavior would indeed require significant effort and potentially
> lead to unforeseen consequences.
> 
> > Unfortunately it took me half an hour just to be able to read through
> > the pieces of this and get far enough I could write a coherent answer
> > to your question. It is a struggle to reply to everything in this level
> > of detail, much as I'd like to :/.
> 
> I'd like to thank you for the time you have invested in this issue! I
> appreciate the help.
> I understand that my proposed solution might not be ideal, but fixing
> the underlying issue in runqueue would be a real challenge.
> If upstream is not interested in this patch, I can also understand that.

It isn't that we're not interested, we just need to make sure something
else doesn't break. A test case to reproduce the issue you're seeing
would help a lot. Ultimately we should perhaps add something to the
tests in meta/lib/oeqa/ to cover the issue.

Cheers,

Richard



  reply	other threads:[~2023-08-04 14:47 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-07-31  9:34 [PATCH] externalsrc: fix dependency chain issues Peter Suti
2023-07-31 10:15 ` [OE-core] " Richard Purdie
2023-07-31 10:43   ` Peter Suti
2023-07-31 11:06     ` Richard Purdie
2023-07-31 12:08       ` Peter Suti
2023-07-31 12:33         ` Richard Purdie
2023-08-02  9:26           ` Peter Suti
2023-08-04 14:47             ` Richard Purdie [this message]
2023-08-07 11:19               ` Peter Suti
2023-08-07 12:55                 ` Richard Purdie
2023-08-07 14:15                   ` Peter Suti
2023-08-08  7:21                     ` Richard Purdie

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=aeed5cf0689586cedccac7d9be2b36d5f5d6c8ad.camel@linuxfoundation.org \
    --to=richard.purdie@linuxfoundation.org \
    --cc=openembedded-core@lists.openembedded.org \
    --cc=peter.suti@streamunlimited.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