Replies: 1 comment 1 reply
-
This sums up mostly how I feel about this. Explicitness of stages makes them predictable and very "standalone" units of operation. I feel it's hard to explain the importance of this sometimes, and I suspect it depends on how often one has had to troubleshoot a build failure from a manifest they didn't themselves create. |
Beta Was this translation helpful? Give feedback.
-
Let's have chat about two osbuild PRs from a philosophy standpoint:
osbuild/osbuild#1506
osbuild/osbuild#1507
Both of these add support for being more 'free' with your stage options; instead of defining the exact kernels we regenerate all kernels, and instead of defining exact paths we define globs for the copy.
I see no technical problem with those approaches as they will still always do the same given the same input tree; however for debugging when reading manifests it's definitely an implicitness I don't like.
Both of these PRs also seem to be geared towards making manual manifest writing easier; but I believe that's something that we discourage in general (?).
For the kernel shenanigans we do have some osbuild-mpp related things that might be able to do this based on the resolved package set however for the glob copies we don't, as those require use to actually construct the tree that is being copied on/from.
Beta Was this translation helpful? Give feedback.
All reactions