Dusty Mabe

some more info.. it appears something with the ext4 filesystem when OSBuild runs the `qemu-img convert` to get from `raw` to `qcow2` is getting corrupted. I converted the image back...

ok further update.. I became increasingly suspicious of the corruption happening earlier in the OSBuild steps themselves rather than when copying the file out of the supermin VM. I decided...

Talked to @jlebon on this. We think (actually more @jlebon, I was just nodding my head to everything he said) that this might be a bug with reflinks on XFS....

@rwmjones in the failure I detailed in https://github.com/coreos/coreos-assembler/issues/3728#issuecomment-1944607900 all that is happening is: - `qemu-img convert input.raw output.qcwo2` - `qemu-img compare -f raw -F qcow2 input.raw output.qcow2` and you can...

> It's extremely suspicious though. Usually qcow2 and qemu-img are rock solid tools. Yes :). I'd like to take a moment here and thank you and all the other maintainers...

OK - I was asked by @sandeen to provide a disk image containing an XFS filesystem with the good and bad generated disk images on it. I had to make...

As a lazy way to test this in other environments I opened a test PR against OSBuild that would then run in their CI - https://github.com/osbuild/osbuild/pull/1594 Sure enough [CI failed](https://gitlab.com/redhat/services/products/image-builder/ci/osbuild/-/pipelines/1178224230)...

Talked with @XanClic and she was able to get the file downloaded

Some more context here about what OSBuild is doing that might be unique here. OSbuild builds things in stages. We have stages that do one or two things, but the...

More data: I patched OSBuild to not `cp` using `--refink=auto` when it copies the first `disk.img` to the second `disk.img` in the workflow described above: ``` diff --git a/stages/org.osbuild.copy b/stages/org.osbuild.copy...