On Thu, 2026-10-15 at 10:46 +0000, Daniel Turull wrote:
> On Thu, 2026-10-15 at 11:16 +0100, Richard Purdie wrote:
> > On Mon, 2026-10-12 at 13:53 +0200, Daniel Turull via lists.openembedded.org 
> > wrote:
> > > From: Daniel Turull <[email protected]>
> > > 
> > > epochfile_read() memoizes its result into __CACHED_SOURCE_DATE_EPOCH
> > > via d.setVar(), so its runtime value leaks into the basehash of any
> > > task that calls it, causing a "metadata is not deterministic" error
> > > once SDE_FILE is populated.
> > > 
> > > Exclude it, since it is not meaningful build metadata.
> > > 
> > > AI-Generated: Uses Kiro with Claude Sonnet 5
> > > Signed-off-by: Daniel Turull <[email protected]>
> > > ---
> > > This should fix the selftest errors reported by Mathieu
> > > ---
> > >  meta/classes-global/base.bbclass | 4 ++++
> > >  1 file changed, 4 insertions(+)
> > > 
> > > diff --git a/meta/classes-global/base.bbclass 
> > > b/meta/classes-global/base.bbclass
> > > index 9553f00432..01a030399e 100644
> > > --- a/meta/classes-global/base.bbclass
> > > +++ b/meta/classes-global/base.bbclass
> > > @@ -249,6 +249,10 @@ do_unpack[postfuncs] += 
> > > "create_source_date_epoch_stamp"
> > >  def get_source_date_epoch_value(d):
> > >      return oe.reproducible.epochfile_read(d.getVar('SDE_FILE'), d)
> > >  
> > > +# epochfile_read() memoizes into this variable via d.setVar(), so its
> > > +# runtime value must not leak into any task's hash.
> > > +oe.reproducible.epochfile_read[vardepsexclude] = 
> > > "__CACHED_SOURCE_DATE_EPOCH"
> > > +
> > >  def get_layers_branch_rev(d):
> > >      revisions = oe.buildcfg.get_layer_revisions(d)
> > >      layers_branch_rev = ["%-20s = \"%s:%s\"" % (r[1], r[2], r[3]) for r 
> > > in revisions]
> > 
> > This is the old syntax and is less maintainable. It can be done
> > alongside the actual function now, for example:
> > 
> > meta/lib/oe/packagedata.py:@bb.parse.vardepsexclude("BB_NUMBER_THREADS")
> > 
> > Could you tweak and resend please?
> 
> Sure. I'll send a v6 on the series. Beside this, is the autobuilder test
> passing?
> 

The autobuilder tests passed, yes. I'd like Joshua's review on the
other patches before we merge though.

Cheers,

Richard
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#247874): 
https://lists.openembedded.org/g/openembedded-core/message/247874
Mute This Topic: https://lists.openembedded.org/mt/121713321/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to