app: Only remount /sysroot if needed

[jlebon]

We should only try to remount /sysroot if we're actually handling the
sysroot repo and the repo isn't writable. We already have public APIs to
check each of those, so let's use them.

Closes: #2485

[kjbracey]

I can confirm this patch fixes my script, applied to 2021.1.

[lucab]

I think this resulted in a change of behavior for a very specific CLI usecase.
rpm-ostree CI does some hackish manipulation via ostree --repo /ostree/foo/bar, which apparently was relying on the previous logic here to remount sysroot as RW.
I'm tying to amend the CI hacks in coreos/rpm-ostree#3246, but I guess it would be nice to special case --repo /ostree/... and --repo /sysroot/ostree/... so that ostree automatically remounts /sysroot.
Or maybe there is a smarter way to do the detection via fstatvfs()?

[jlebon]

I'm tying to amend the CI hacks in coreos/rpm-ostree#3246, but I guess it would be nice to special case --repo /ostree/... and --repo /sysroot/... so that ostree automatically remounts /sysroot. Or maybe there is a smarter way to do the detection via vstatvfs()?

Hmm, maybe. I think this is somewhat specific to the rpm-ostree CI which likes to keep test repos in /ostree/repo/ to ensure hardlinks with the sysroot repo. Maybe simplest is to have a hidden stamp file in /run somewhere and we just always remount rw in that case in this code? That way in the rpm-ostree CI, we'd just need a touch /run/... upfront.

Though if the only fallout is coreos/rpm-ostree#3246, then I'd vote for not doing anything at all for now. :)

[lucab]

Yes I haven't seen any other fallout, at least yet.
I'm not super happy with the proliferation of magic stamp files as reliable interfaces for consumers.
Thinking a bit more about my previous copmment, the statfs approach would probably be to broad.
The rpm-ostree got easily fixed already.
Overall, I agree on:

• keeping things as is going forward
• if we notice further breakage for similar cases, just special-casing the /ostree path-prefixes