Composefs changes, disable failing ostree test #2277
Conversation
We were writing out the ImageId in output to `bootc status` instead of writing the manifest digest Fixes: bootc-dev/bink#90 Signed-off-by: Pragyan Poudyal <pragyanpoudyal41999@gmail.com>
Johan-Liebert1
force-pushed
the
cfs-status-fix
branch
from
856e385 to
17f1a1d
Compare
| Ok(img_conf) | ||
| // Compute the manifest digest | ||
| let mut hasher = Sha256::new(); | ||
| hasher.update(&serde_json::to_vec(&img_conf.manifest).context("Serializing image manifest")?); |
There was a problem hiding this comment.
This is re-serializing the manifest object. But for our purposes here, we want to be certain we end up with the same manifest digest as the remote. Are we guaranteed a match here? Ideally we'd just hash the raw manifest string we originally fetched from the remote at pull time, which IIRC is how this works on the ostree backend.
There was a problem hiding this comment.
This is more "legacy" code for backwards compatibility for older systems where we used to store the config + manifest in an .imginfo file. To answer your question, yes, it should, in theory, produce the same digest since we pretty much used to store the manifest and config unchanged. Now though, we do what ostree does and store the manifest digest string inside the .origin file
There was a problem hiding this comment.
To answer your question, yes, it should, in theory, produce the same digest since we pretty much used to store the manifest and config unchanged.
I think the point Jonathan is making here is that we should hash the raw data and not re-serialize.
But @Johan-Liebert1 is also right in this is just the legacy path, which I think we can schedule to remove at some point and isn't really relevant anymore.
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
cap_std's `metadata_optional` follows symlinks and errors out if the symlink target is outside the current root Use `symlink_metadata_optional` when checking for removed files in new etc so that symlinks that resolve to a path outside of new_etc don't get followed Fixes: bootc-dev#2278 Signed-off-by: Pragyan Poudyal <pragyanpoudyal41999@gmail.com>
This test ends up failing for systems with ostree v2026.2. Disable it until we have a fix Signed-off-by: Pragyan Poudyal <pragyanpoudyal41999@gmail.com>
Johan-Liebert1
changed the title
Enable the rhcontainerbot/bootc COPR repo before building the rootfs so that the composefs imageDigest fix from bootc-dev/bootc#2277 gets picked up once it merges and COPR builds resume. Signed-off-by: HarshwardhanPatil07 <harshpat@redhat.com>
3 participants
We were writing out the ImageId in output to
bootc statusinstead of writing the manifest digestFixes: bootc-dev/bink#90
etc-merge: Use
symlink_metadata_optionalcap_std's
metadata_optionalfollows symlinks and errors out if thesymlink target is outside the current root
Use
symlink_metadata_optionalwhen checking for removed files in newetc so that symlinks that resolve to a path outside of new_etc don't
get followed
Fixes: #2278
tests: Skip test-loader-entries-source for ostree v2026.2
This test ends up failing for systems with ostree v2026.2. Disable it
until we have a fix