Thanks for the feedback, Bruce. It’s likely not a small task, so it’ll probably be a bit before I can get the time allotted to work on it, but I think it’s something that we’d definitely like to pursue. I’ll reach out again for feedback when we have the capacity to staff this and have done some initial work on putting together ideas and sketching a rough proposal and proof of concept.
Much obliged! -Sean McKay From: Bruce Ashfield<mailto:[email protected]> Sent: Tuesday, September 29, 2020 4:40 AM To: McKay, Sean<mailto:[email protected]> Cc: [email protected]<mailto:[email protected]> Subject: Re: [linux-yocto] Interim task between compile/bundle_initramfs On Mon, Sep 28, 2020 at 6:45 PM Sean McKay <[email protected]> wrote: > > Hi Bruce, et al. > > Has anyone given any thought to trying to add an sstate enabled task between > compile and bundle_initramfs that would capture the necessary artifacts for > the downstream tasks to run without necessarily actually recompiling the > kernel if nothing in the kernel code/config has changed? > I ask because we’ve got a few embedded systems that use just a bundled > kernel/initramfs to run, and it seems quite a waste of compute time to > recompile the kernel every time the initramfs changes image signature (and we > can’t be the only ones for whom the compute time savings would be nonzero). > > > On the other hand, I’m guessing smarter people than I have already thought of > said question… so before I dig too far, I figured I’d ask if there’s a reason > this hasn’t been done, or if you think it’s worth pursuing (and no one else > has had the time yet)? It has been thought about, but I've never seen a complete enough solution to it to consider for merging. That being said, some of the core features that have added/fixed to bitbake/oe in the last year or so, should help with this (everything from sstate fixes to hash equiv to multiconfig). Fundamentally you are talking about a task to re-pack the initramfs based on (sstate) artifacts. This is something that could have been forced/done in a custom class based on deployed artifacts only, but using sstate would catch cases where things did change, and you really shouldn't be using the deployed variants. I think it is worth pursuing, but I'm sure there are corner cases I'm not catching off the top of my head. I'd be interested in helping with any POCs or design elements (or implementation), etc. So feel free to fire things off to me or the list as you need. Bruce > > Thanks! > > -Sean McKay > > > > -- - Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end - "Use the force Harry" - Gandalf, Star Trek II
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#9099): https://lists.yoctoproject.org/g/linux-yocto/message/9099 Mute This Topic: https://lists.yoctoproject.org/mt/77185489/21656 Group Owner: [email protected] Unsubscribe: https://lists.yoctoproject.org/g/linux-yocto/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
