Add new plugin slots for pre/post chroot events #2669
Conversation
|
Looks good and pretty straight forward. May be should get a test case. It might also be a good idea to point out in the docs that these are called even when running without --root. Technically we omit the chroot call in this case but the hooks are ofc executed no matter what (as they should be) |
Yep. I actually have a test-case written in the form of "plugin development" test, but it needs some cmake infra bits first. But, this PR isn't in any particular hurry.
Um, no they don't. rpmChrootIn() and rpmChrootOut() do always get called but they return with a no-op from the first condition. |
These pre/post hooks get called before and after every chroot() call rpm makes. The initial inspiration came from the now removed experimental non-privileged chroot support with unshare() but this is probably an interesting boundary for various other purposes too. Adding a testcase is annoyingly complicated as we need to install into an empty root for this to trigger. And in the empty chroot have no shell, and because this is a real empty chroot we can't just fake it either. We could use a -p <lua> scriptlet but then that wouldn't test scriptlet_fork_post because Lua scriptlets aren't currently forked. So we need two test-cases anyhow, just reuse the same case with --nodeps and --noscripts.
|
And now with a testcase. |
|
The more I sleep on this, the less convinced I become that this is a good idea. Quite the contrary actually. "Uneasy" is the word. That instinct is rarely wrong, so I'm canceling this PR. If at some point an actual use-case presents itself, we can always re-evaluate. |
These pre/post hooks get called before and after every chroot() call rpm makes. The initial inspiration came from the now removed experimental non-privileged chroot support with unshare() but this is probably an interesting boundary for various other purposes too.