Eighth_Doctor | slyon: hey, you around? | 14:48 |
---|---|---|
slyon | Eighth_Doctor: hey! | 14:49 |
Eighth_Doctor | how are we doing for https://github.com/canonical/netplan/pull/285? | 14:50 |
Eighth_Doctor | and VXLAN support? | 14:50 |
Eighth_Doctor | also, is it intentional that netplan can't built on RHEL 8? | 14:51 |
slyon | Eighth_Doctor: Implementation is done for VRF, currently under review. | 14:51 |
Eighth_Doctor | I tried building netplan on RHEL 8, and it choked on a missing glib2 function | 14:51 |
slyon | VXLAN is next up on my list, a draft has been posted in https://github.com/canonical/netplan/pull/272 – but that needs quite some extra work, still | 14:52 |
slyon | both VRF and VXLAN are planed to be released in the upcoming netplan v0.105 due in August | 14:52 |
Eighth_Doctor | okay | 14:52 |
slyon | It is certainly not intentional that netplan can't be built on RHEL 8... did you try using `make` or `meson`? | 14:53 |
slyon | ah the breaking glib change... I think I made it functional using old and new glib... let me check | 14:53 |
slyon | Eighth_Doctor: this patch should be included in the current 0.104 release: https://github.com/canonical/netplan/commit/15c98fec3948c19854f270225f552aa50d9cabcd | 14:54 |
slyon | is there any other glib problem? | 14:54 |
Eighth_Doctor | yeah | 14:55 |
Eighth_Doctor | gimme a sec and I'll pull it up | 14:55 |
Eighth_Doctor | I do vastly prefer meson over make, but I haven't switched the Fedora/EPEL package to it yet | 14:56 |
Eighth_Doctor | I don't think 0.104 had it | 14:56 |
slyon | Eighth_Doctor: ok. for < 0.104 you could probably apply/adopt the patch I pasted above to make it work, using old glib | 14:57 |
Eighth_Doctor | the glib patch is in 0.104 | 14:58 |
Eighth_Doctor | it still ftbfs | 14:58 |
Eighth_Doctor | there's a different glib2 function | 14:58 |
* Eighth_Doctor is running the build again to get that issue | 14:58 | |
slyon | wrt meson: we need a relatively new meson (>= 0.61 IIRC) so that probably won't work for RHEL | 14:59 |
Eighth_Doctor | yeah, it's currently at 0.58 | 14:59 |
Eighth_Doctor | so I think I'm stuck with make for a while | 15:00 |
slyon | yes. | 15:00 |
Eighth_Doctor | slyon: g_clear_list() doesn't exist in glib2 2.56 | 15:00 |
slyon | let's see if we can find a solution to that other glib issue... I might have missed it in Debian/Ubuntu | 15:00 |
Eighth_Doctor | which is what's present in RHEL 8 | 15:00 |
slyon | Eighth_Doctor: I need to run for a meeting now. Will have a look at g_clear_list() later. (Or if you feel like, you could prepare a patch/PR) | 15:01 |
slyon | thank you very much for your involvement in netplan to cover the RPM side, btw! | 15:01 |
Eighth_Doctor | 👍️ | 15:02 |
Eighth_Doctor | slyon: I'm not sure how to deal with g_clear_list(), since I am not very familiar with glib2 stuff | 15:08 |
Eighth_Doctor | but if you can find a fix later today, that'd be great | 15:08 |
slyon | I guess we could add something like this https://gitlab.gnome.org/GNOME/glib/-/issues/1943 to a header, isolated by a "#if GLIB_CHECK_VERSION (2, 56, 0)" macro to locally define that function for builds on old glib | 15:33 |
slyon | Eighth_Doctor: does something like this work for you? https://pastebin.com/NtdhQPHH | 15:55 |
Eighth_Doctor | slyon: I can't access pastebin.com here | 15:55 |
Eighth_Doctor | can you use paste.centos.org or similar pastebin? | 15:55 |
slyon | https://paste.centos.org/view/acceab72 | 15:56 |
slyon | trying to reproduce this on Ubuntu Bionic (glib 2.56) I can see another similar issue: g_hash_table_steal_extended -Werror=implicit-function-declaration – It might need a similar fix | 15:56 |
Eighth_Doctor | 👍️ | 15:57 |
Eighth_Doctor | yeah, we might want to introduce CI on CentOS 8 once this is all fixed | 15:57 |
slyon | This has been in glib since 2018 times... so I'm not sure if I really want to have glib symbols defined in upstream netplan, just for backwards compatibility. Maybe it'd be better to add that patch to the rpm package locally? | 15:58 |
slyon | It would be great to get CentOS 8 CI, though. (I want Debian, too, but didn't find the time to set that up, yet) | 15:58 |
Eighth_Doctor | well, at least having CentOS 8 in CI will give decent compatibility coverage | 15:59 |
slyon | ack | 15:59 |
Eighth_Doctor | my recommendation is to use alternate names rather than keeping the glib names for compat functions | 16:03 |
Eighth_Doctor | e.g. np_g_* rather than just g_* | 16:03 |
slyon | yeah.. but then we'd need to differentiate between the two symbol names when calling the funciton... As I said, I'm not sure if id' want to include those glib functions inside netplan headers at all... :-/ I need to think about this a bit more. | 16:09 |
slyon | gtg now | 16:09 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!