I've reviewed the package in the jammy queue, since Chris seemed to only get to the lunar one in his SRU shift. I have some questions that I expect may apply to both uploads.
It was initially confusing to me that this was a backport of a version that's been superseded in mantic, and also that it includes the addition of a number of files under debian/patches; I wondered why these had not been committed upstream. Then I saw that these were present in the -7 package, so this is ok from that perspective.
So then I diffed the upload in the jammy queue against 0.106.1-7 and found other packaging differences that don't make sense:
Why is this backport missing symbols entries that are present in the unstable version? I see you've included a patch, debian/patches/sru-compat/0024-Avoid-unnecessary-export-of-API.patch, that patches these symbols out (which, confusingly accomplishes this by removing the NETPLAN_INTERNAL tag, which on its surface would appear to do the opposite?). But the net effect here is that you are now introducing a libnetplan0 package in the archive, whose version number is >= 0.106, but whose ABI does not match what is declared in later versions - so if someone installs this package in jammy, then upgrades to mantic, apt may not have correct information about the necessary order in which packages must be upgraded, because a new package using these symbols in mantic will think that the version in jammy satisfies the dependency!
So I think the correct course of action here is to NOT patch these out of the library's ABI.
I've reviewed the package in the jammy queue, since Chris seemed to only get to the lunar one in his SRU shift. I have some questions that I expect may apply to both uploads.
It was initially confusing to me that this was a backport of a version that's been superseded in mantic, and also that it includes the addition of a number of files under debian/patches; I wondered why these had not been committed upstream. Then I saw that these were present in the -7 package, so this is ok from that perspective.
So then I diffed the upload in the jammy queue against 0.106.1-7 and found other packaging differences that don't make sense:
--- netplan. io-0.106. 1/debian/ libnetplan0. symbols 2023-07-20 06:11:50.000000000 -0700 io-0.106. 1/debian/ libnetplan0. symbols 2023-08-24 03:11:14.000000000 -0700 ed_address@ Base 0.104 backend@ Base 0.104 present@ Base 0.106 rule_present@ Base 0.106.1 in_array@ Base 0.106.1 key@Base 0.100 ordered@ Base 0.99
+++ netplan.
@@ -26,9 +26,6 @@
get_unspecifi
global_
global_state@Base 0.104
- is_route_
- is_route_
- is_string_
is_wireguard_
netdefs@Base 0.99
netdefs_
Why is this backport missing symbols entries that are present in the unstable version? I see you've included a patch, debian/ patches/ sru-compat/ 0024-Avoid- unnecessary- export- of-API. patch, that patches these symbols out (which, confusingly accomplishes this by removing the NETPLAN_INTERNAL tag, which on its surface would appear to do the opposite?). But the net effect here is that you are now introducing a libnetplan0 package in the archive, whose version number is >= 0.106, but whose ABI does not match what is declared in later versions - so if someone installs this package in jammy, then upgrades to mantic, apt may not have correct information about the necessary order in which packages must be upgraded, because a new package using these symbols in mantic will think that the version in jammy satisfies the dependency!
So I think the correct course of action here is to NOT patch these out of the library's ABI.
--- netplan. io-0.106. 1/debian/ netplan. io.install 2023-07-25 03:18:01.000000000 -0700 io-0.106. 1/debian/ netplan. io.install 2023-08-24 01:24:24.000000000 -0700 system- generators/ netplan netplan/ *
+++ netplan.
@@ -1,3 +1,4 @@
+lib/netplan/*
lib/systemd/
usr/libexec/
usr/share/*
Why is this present only in the jammy backport and not in the unstable version? (Particularly as /lib should be avoided now in favor of /usr/lib)