-
Notifications
You must be signed in to change notification settings - Fork 11
Update kernel config after make oldconfig
#207
Comments
@jodh-intel I'm a bit doubtful of getting the kernel .config file from the buildroot of a certain distro build, considering the make version stuff. |
@erick0z - hmm, yes, there is a potential trust issue there too I guess. Your plan sounds good to me, but could you comment @grahamwhaley? |
I'm a touch confused - I don't quite get the bit about 'getting the kernel .config from the buildroot of a certain distro build'. If you mean 'take the kernel config from the distro we are building on' then:
So, (and I suspect this aligns with what is being said above), I think we:
Does that make sense? |
@grahamwhaley Currently, we are not getting the kernel .config from any buildroot and I guess it will never happen. @jodh-intel just mentioned in the first comment. It happens that the kernel .config that lives in the packaging repo is not the same shipped in the packages. I mean, in the buildroot a new one is generated based on the packaging one using |
AFAIK, packaging stores the official kernel config and patches for CC. So, if the kernel is updated (normally, with a PR), it must contain the updated config as well. |
Now I see where my confusion came from. I was looking at these two package templates which copy the config file before running
... but I've just realised that the In other words, the initial kernel configuration right now (aiui) is: ... but that file is different to
|
Strike that - I was looking at stale info from https://github.com/clearlinux-pkgs/linux-container/blob/master/config. The configs are almost identical and #212 has just been raised that will make them identical. |
Hi @erick0z - I'm still a bit confused: Both
|
The packaging script currently copy the kernel config into the build tree and run
make oldconfig
. But the update.config
file that is generated is not committed back to this repository. This is confusing as shown here.We should update the packaging script to commit the changes along with a message in the commit that includes:
make oldconfig
was run on (shouldn't matter, but just in case).I'd also consider adding:
make --version | head -1
as GNU Make has been known to contain bugs ;)However, since
make oldconfig
may change not just the comment header but potentially the value of config options (even add/remove them), we should run CI tests on the generated config to ensure the config results in packages that produce a usable system. We should probably also consider the PnP perspective of any changes too (/cc @grahamwhaley).See also:
The text was updated successfully, but these errors were encountered: