On Mon, 2023-10-23 at 11:17 +0200, Alexander Kanavin wrote:
> On Sun, 22 Oct 2023 at 13:55, Richard Purdie
> <[email protected]> wrote:
> > > > Seems like that ('auto') had no effect. :-(
> > > 
> > > Hit send too quickly. Should I try with your suggestion?
> > 
> > It is worth a shot, yes. One causes a local hashequiv whereas the other
> > disables it entirely.
> 
> Seems like it worked:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/127/builds/2314/steps/15/logs/stdio

Great!

> Should I fold BB_SIGNATURE_HANDLER = "OEBasicHash" into this commit,
> with a FIXME perhaps? Tools should work without special tweaks, so it
> needs to be addressed.

We should add the workaround into the commit and leave a note, yes.

Hashequiv is 'fun' since it means in a new build of something, the
build can track onto an older build/hash chain depending on whether it
matches it or not.

I've not looked into the test specifically and you're right, we want
people to be able to get diffs reliably regardless of whether it is
enabled or not. Whether the issue is with the test case or the tooling,
I'm not really sure. I suggested it as it was the most likely source of
the issues and definitely worth ruling out if it wasn't.

Note that the armhost case is interesting since the hash equivalence
mapping can be present but the sstate might not be in the native case
if it was built only on x86.

Cheers,

Richard
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#189617): 
https://lists.openembedded.org/g/openembedded-core/message/189617
Mute This Topic: https://lists.openembedded.org/mt/102035446/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to