| LeePen | Centurion_Dan: yes but surely the library loader will complain when it doesn't find sd_journal_* symbols for a binary it is loading? | 11:38 |
|---|---|---|
| Centurion_Dan | only if the program uses those symbols. | 11:39 |
| LeePen | I am just looking for a concrete example on my VM | 11:41 |
| Centurion_Dan | We could of course ask the elogind developers to stub out as NO_OP the api for sd_journal and the other api bits they're missing. | 11:45 |
| LeePen | So on my ceres VM I have /usr/lib/x86_64/publseaudio/libpulsecommon-12.2.so. It comes from a debian merged package and has an undefined symbol | 11:46 |
| LeePen | 0000000000000000 DF *UND*0000000000000000 LIBSYSTEMD_209 sd_journal_send | 11:46 |
| LeePen | 11:46 | |
| LeePen | sd_journal_send is only in libsystemd0 | 11:47 |
| Centurion_Dan | but we have to rebuild pulseaudio anyway... | 11:48 |
| LeePen | Yes, I know, so when we build our own we can build against libelogind in the first place and all will be fine | 11:49 |
| LeePen | I thought you were wondering about using binaries build against libsystemd with libelogind? | 11:49 |
| Centurion_Dan | or just fix libelogind to include all the symbols... and maybe map sd_journal_send to a standard syslog call. | 11:54 |
| LeePen | Yes, I can see that might work, but I have my doubts that elogind upstream would be interested in that. I don't think it is their motivation for having elogind. | 11:56 |
| LeePen | They have a pure elogind system and everything is built against it's headers, links with libelogind and works. | 11:57 |
| LeePen | We are stuck in the middle with some packages from Debian built against systemd headers and trying to use them on libelogind systems. | 11:58 |
| LeePen | I know we don't want to fork more packages than is necessary, but I wonder if we might have to. | 11:59 |
| LeePen | I haven't looked at amprolla at all, so this is no more than an idle musing, but might it be possible to arrange the recompilation of libsystemd rdepends semi-automatically to avoid too much manual work? | 12:01 |
| Centurion_Dan | i've commented on the original thread about this on elogind's page. We'll see how favorable they are to the idea. | 12:30 |
| LeePen | Centurion_Dan: OK, great. | 13:59 |
| Centurion_Dan | LeePen: does pulseaudio in your test case fail to run, or just complain loudly? | 14:13 |
| Centurion_Dan | LeePen: https://github.com/elogind/elogind/issues/97 | 14:30 |
| fsmithred | if I reply to a message on devuan-bugs list, does it also go to the bug report at bugs.devuan.org, or do I also need to reply to bug#@bugs.devuan.org? | 17:02 |
| LeePen | Centurion_Dan: Yes, /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-12.2.so: symbol sd_journal_send version LIBSYSTEMD_209 not defined in file libsystemd.so.0 with link time reference | 17:12 |
| LeePen | 17:12 | |
| LeePen | Well, actually I think that is the linker complaining and then exiting without running the binary. | 17:19 |
| LeePen | Centurion_Dan: I see Yamakuzure is happy to add the missing symbols to libelogind. Good news! | 17:35 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!