| bb|hcb | bgstack15: "The exact contents of the package from Ceres should go in Chimaera." - what do you mean? Actually above I am wrong, because this is not a backport but a stable update, so the changes from 0.0.2 should be merged in the suites/chimaera branch and the version should be 0.0.1-1+deb11u1 | 00:04 |
|---|---|---|
| bgstack15 | The contents of systemctl-service-shim package should be identical across Ceres and Chimaera. Doesn't that mean the package version should be the same? | 00:20 |
| bgstack15 | I'm not sure I've dealt with a package across multiple releases yet. | 00:20 |
| bgstack15 | Version 0.0.1 is broken, so version 0.0.2 should be run everywhere. | 00:21 |
| bb|hcb | The point is, that a package with the same version will not be replaced on upgrade. That is why stable updates are always based on the previous version but with suffix that ensures that the update version sorts after the existing one. The contents are another story, in this case it is perfectly OK to merge all changes, so practically the contents would be identical to 0.0.2-1 but the version | 00:25 |
| bb|hcb | will be 0.0.1-1+deb11u1 | 00:25 |
| jjakob | shouldn't /bin/systemctl also be here https://git.devuan.org/devuan/systemctl-service-shim/src/branch/master/debian/systemctl-service-shim.links | 00:26 |
| bb|hcb | jjakob: check the suites/unstable branch | 00:27 |
| jjakob | bb|hcb: yes? that file is the same | 00:28 |
| jjakob | what am I supposed to see | 00:28 |
| bb|hcb | jjakob: I am not aware with that package, you are right master branch is identical to suites/unstable | 00:30 |
| jjakob | this is the package bgstack15 is also asking about | 00:31 |
| bgstack15 | It is very important for the freeipa use case that systemctl NOT be in /usr/bin/ | 00:32 |
| jjakob | I'm not saying that | 00:33 |
| jjakob | I'm saying it should be in /bin | 00:33 |
| jjakob | bb|hcb: how would you then know that 0.0.1-1+deb11u1 is actually the same as 0.0.2-1 ? | 00:34 |
| jjakob | bgstack15: /usr/lib/python3/dist-packages/ipaplatform/base/paths.py: SYSTEMCTL = "/bin/systemctl" | 00:34 |
| jjakob | it needs it to be in /bin, you see | 00:35 |
| bgstack15 | Hm, my processes for a new system work as-is. Perhaps I am making that symlink somewhere. I will need to build a new VM and test joining it to my domain. | 00:35 |
| jjakob | it did not work for me | 00:36 |
| jjakob | I installed freeipa-client and tried ipa-client-install, which gave me the file or directory not found: /bin/systemctl error | 00:36 |
| jjakob | after I created the symlink by hand, I ran into that infinite loop bug that is fixed in 0.0.2 | 00:37 |
| bgstack15 | Yeah, so definitely we need to work with >= 0.0.2 and forget 0.0.1 entirely. | 00:37 |
| bb|hcb | jjakob: By the message in changelog... No other way | 00:42 |
| jjakob | that seems hard. is there no way to put the upstream version in there somehow | 00:43 |
| bb|hcb | That is an option, to use version like 0.0.1-1+deb11u0.0.2-1 | 00:44 |
| bgstack15 | I don't understand. Why can we not simply put the version number 0.0.2 (since it appears that just a packaging-fix will suffice) in Chimaera? | 00:48 |
| bgstack15 | Is the version number not allowed to change at all? | 00:48 |
| bb|hcb | Think about the upgrade path - after a user upgrades to deadalus, the package will not be upgraded. That is why versions in chimaera should be smaller than the ones in daedalus | 00:55 |
| bb|hcb | In systemctl-service-shim | 00:57 |
| bb|hcb | that would not be a problem because it is a shell script, but it is better to follow the best practice | 00:57 |
| bb|hcb | For compiled binary packages, especially ones that rely on shared libs that can cause a real trouble and break stuff - the chimaera one should be built in chimaera chroot, also have a smaller version, in order after upgrade to switch to the daedalus package that is built on daedalus... | 01:00 |
| jjakob | isn't that why the "deb10" part exists - it will then be deb11 ? or is that part not necessarily present? | 01:05 |
| jjakob | bgstack15: alternatively patch python3-ipalib paths.py to change the hardcoded path, if you think creating that /bin/systemctl symlink would be a problem for something | 01:06 |
| bgstack15 | Doing anything with systemctl-service-shim is going to be easier than maintaining a patch on top of freeipa. | 01:12 |
| bgstack15 | bb|hcb: thank you for explaining that reasoning. I don't use any non-rolling Devuan installs, but surely version numbers get incremented in stable releases without all that versioning rigamarole, right? | 01:13 |
| bb|hcb | jjakob: deb10 is stable update for debian 10 (buster), deb11 for bullseye, etc. As far as I am aware there is no convention in Devuan for stable updates of native packages, but it would be best to wait for LeePen to confirm | 01:26 |
| fsmithred | 0.0.2-1~bpo...? | 01:27 |
| bb|hcb | bgstack15: yw :) they should be, else the upgrade path will break and stable/oldstable users will not be able to properly upgrade | 01:29 |
| bb|hcb | fsmithred: no, bpo comes from backports.org and is intended for backports (they are optional), stable updates are different and included in the stable or oldstable | 01:29 |
| bb|hcb | Initially I also got confused that bgstack15 is doing a backport, while that is a stable update | 01:30 |
| fsmithred | yeah, I see that now | 01:30 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!