On Wed, 2023-05-03 at 14:42 +0200, Alexander Kanavin wrote: > On Wed, 3 May 2023 at 14:39, Michael Opdenacker via > lists.openembedded.org > <[email protected]> wrote: > > Then, we I ran my tests on a fresh Fedora container, bitbake didn't warn > > me but "xz" was missing, but this prevented the uninative shim from > > being installed. It seems that at least on Fedora, "xz" isn't pulled by > > the other host tools we install. > > > > I guess most people has "xz" anyway, which could explain why this wasn't > > reported before. > > HOSTTOOLS is used to filter tools from the host in $PATH, at least in > bitbake tasks, so is uninative extractor bypassing that somehow then?
The uninative code calls tar from the host. It then probably depends upon how tar has been configured and whether it calls xz or uses a library? I'm guessing in the context of uninative, PATH hasn't been locked down yet. I think we probably should add xz to HOSTTOOLS at this point but it probably does also mean we should change ASSUME_PROVIDED and there are some things with an xz dependency which will need to use xz- replacement-native since they probably need library components and not just xz itself. Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#180825): https://lists.openembedded.org/g/openembedded-core/message/180825 Mute This Topic: https://lists.openembedded.org/mt/98658155/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
