Hi Ross,
I did that.
current behavior:
$ devtool modify busybox
Initialising tasks: 100% |#######################################|
Time: 0:00:00
Sstate summary: Wanted 14 Local 0 Mirrors 0 Missed 14 Current 6 (0%
match, 30% complete)
NOTE: Executing Tasks
...
your new behavior:
@@ -52,7 +52,7 @@ def check(args):
try:
output = subprocess.check_output(
- 'bitbake -n %s' % ' '.join(args.target),
+ 'bitbake --runall build -n %s' % '
'.join(args.target),
stderr=subprocess.STDOUT,
env=env,
$ devtool modify busybox
Initialising tasks: 100% |#######################################|
Time: 0:00:00
Sstate summary: Wanted 14 Local 4 Mirrors 0 Missed 10 Current 6 (28%
match, 50% complete)
NOTE: Executing Tasks
...
It just start fetching, patching... tasks without any dependencies
build. Looks like, that this is exactly what we wanted to achieve.
Let me ask the question about the percentage... why is 28% of match?
Regards,
Andrej
On Thu, 2022-10-13 at 10:15 +0000, Ross Burton wrote:
>
>
> > On 13 Oct 2022, at 09:23, Andrej Valek via lists.openembedded.org
> > <[email protected]> wrote:
> > I had some time and made some more testing.
> >
> > Looks like, that the problematic commits are here:
> > https://github.com/openembedded/openembedded-core/commit/41d7f1aa2cc9ef5dba4db38435402d4c9c0a63e1
> > https://github.com/openembedded/openembedded-core/commit/6e2cbfc561dac89bf9183d24d90e52f7d9117826
> > https://github.com/openembedded/openembedded-core/commit/030d56e2e8ece93472adc51fe467221d846c9ac0
> >
> > Means, that recursive do_build task is not being executed on image
> > level anymore. After re-creating/adding the meta class and inheriting
> > it in populate_sdk_base, the image now contains all the missing deps.
> >
> > I tried to execute the "bitbake my-image -c populate_sdk_ext --runall
> > build", but "--runall" destroyed the populate_sdk_ext task, so in
> > this
> > case is unusable.
> >
> > So what's the right way to get it working again?
>
> So the good news is that a few others reported this issue and a bug was
> filed:
>
> https://bugzilla.yoctoproject.org/show_bug.cgi?id=14626
>
> Annoyingly, nobody who saw the bug remembered this thread.
>
> My hunch was that dependency rationalisation meant less
> do_populate_sysroot calls, based on how the eSDK creation is done. If
> you’ve isolated it down to those commits then that’s an indication my
> hunch was right.
>
> Can you test something for me? Switch back to pristine oe-core without
> the reverts, and add the ‘—runall build’ option to the bitbake call in
> scripts/oe-check-sstate (line ~54).
>
> Thanks,
> Ross
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#171701):
https://lists.openembedded.org/g/openembedded-core/message/171701
Mute This Topic: https://lists.openembedded.org/mt/92019337/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
