**** BEGIN LOGGING AT Tue Mar 01 02:59:56 2022 Mar 01 07:44:49 good morning Mar 01 15:03:12 hmm, do_populate_sdk fails because the postinstall scriptlet of pulseaudio fails. It seems to fail because it can't find pulseaudio.service. But I obviously don't need anything to do with the pulseaudio systemd service file in the SDK Mar 01 15:09:08 removing it with TOOLCHAIN_HOST_TASK_remove = "pulseaudio" and TOOLCHAIN_TARGET_TASK_remove = "pulseaudio" doesn't help Mar 01 15:19:53 is there a way to get pulseaudio to not run its postinstall script, but only when being installed to the SDK? Mar 01 15:25:40 mort: what branch/release are you using? And are you using a BSP layer? I don't see anything in master that would lead to that, and I'm pretty sure SDK generation is part of CI with poky.conf, which has pulseaudio enabled Mar 01 15:28:19 hmm we have some stuff in a pulseaudio bbappend file, it creates a pulseaudio service with `inherit systemd` and `SYSTEMD_SERVICE_${PN} = " pulseaudio.service" but I can't see anything weird Mar 01 15:28:47 https://p.mort.coffee/t9V.sh this is the bbappend Mar 01 15:32:07 https://p.mort.coffee/vQ0.txt#5307-5329 here's the log file from it, I don't see any connection between that and the content in the bbappend Mar 01 15:32:15 we're on gatesgarth btw, I know that's eol but it's what we have for now Mar 01 15:34:12 I'm a bit puzzled, you seem to duplicate some of what I'd have thought the systemd support in the base recipe in poky is doing Mar 01 15:35:43 indeed, the inherit systemd and SYSTEMD_SERVICE_${PN} stuff is in the original recipe too Mar 01 15:35:47 not quite sure why that's there Mar 01 15:36:02 same with the user stuff Mar 01 15:36:13 wrt the failure, I've no good idea other than to perhaps dig down into the workdir and see what the postinstall scriptlet looks like (in the .spec file if you're using rpm, I believe) Mar 01 15:36:41 and compare against what it contains in a stock build with poky, maybe Mar 01 15:37:50 is there some way to just exclude pulseaudio? I don't exactly need a pulseaudio server as part of the sdk, just the client libraries Mar 01 15:38:09 I'd guess it's the systemd_postinst from systemd.bbclass that fails, but it'd take some poking around to see why Mar 01 15:38:54 mort: perhaps try your TOOLCHAIN_TARGET_TASK_remove with "pulseaudio-server"? Mar 01 15:39:33 I'll try that Mar 01 15:50:37 smurray: that didn't help, sadly Mar 01 15:53:20 https://p.mort.coffee/u6D.txt#4654-4684 this seems like the postinst script which fails Mar 01 15:53:42 I assume Mar 01 16:00:24 mort: that's definitely the systemd_postinst bit from systemd.bbclass. Perhaps someone else has an idea of how to avoid triggering it Mar 01 16:03:15 mort: you might want to see if just relying on the SYSTEMD_SERVICE_${PN}-server from the recipe in poky works versus your addition of SYSTEMD_SERVICE_${PN} in your bbappend Mar 01 16:22:10 smurray: would you look at that, removing that SYSTEMD_SERVICE_${PN} did the trick Mar 01 16:23:28 I still don't understand why that would break it only when running the postinst for the sdk and not when running the postinst for the actual target, but there's probably some good reason Mar 01 16:25:33 mort: you'd have to look at the packages making it into the SDK, I suspect pulseaudio-server does not Mar 01 16:26:00 sounds plausible Mar 01 16:26:06 mort: while pulseaudio would because I imagine pulseaudio-dev depends on pulseaudio (as that's the default behavior) Mar 01 19:19:32 mort: Here's the longer explanation of what's going on with your pulseaudio problem. Normally when you add a service file to SYSTEMD_SERVICE_*, the code in systemd.bbclass will verify that it exists (so there are no problems later when the package is installed). This happens before the installed files are split per package. The problem in your case is that you have added the same unit file to both SYSTEMD_SERVICE_${PN} and Mar 01 19:19:32 SYSTEMD_SERVICE_${PN}-server. This will fool the code in systemd.bbclass since while it will check for the file's existence for both packages, both tests will find the same file. However, when the packaging actually happens, the file will obviously only end up in one of the packages, and the other package will be broken. **** ENDING LOGGING AT Wed Mar 02 02:59:56 2022