| Mister_Magister | it might not be merged but i hope talks start | 00:00 |
|---|---|---|
| Mister_Magister | i'm very proud of myself for implementing those gestures :3 | 00:00 |
| Mister_Magister | okay now for a tricky question, how do i make it so that sensorfw from my vendor is pulled instead of jolla's | 00:09 |
| Mister_Magister | ig all I can do is to stay version in front of it i guess | 00:09 |
| nightishaman | okay, strace has executed | 00:15 |
| nightishaman | lets look at it | 00:15 |
| nightishaman | mal: it mounts twice, first it mounts /linkerconfig to a tmpfs, then it mounts /linkerconfig/default to /linkerconfig | 00:18 |
| nightishaman | according to the strace | 00:18 |
| Mister_Magister | mal: do you mind explaining to me, why sensorfw package, which is version 0.14.4, in jolla repos is 0.15.0-1.14.1.jolla? Where the hell did 0.15.0 come from? | 00:18 |
| Mister_Magister | it'sa freaking tag | 00:19 |
| Mister_Magister | tag not matching a version should be criminal offence | 00:20 |
| Mister_Magister | pushed my tag 1.0 so that it's always above jolla's no matter what | 00:22 |
| mal | Mister_Magister: someone forgot to update the version | 00:31 |
| Mister_Magister | :) | 00:31 |
| Mister_Magister | lemme make a PR i'm sure it will be merged in timely fasion :D | 00:32 |
| nightishaman | builtins.cpp:1227 | 00:32 |
| mal | but if you build that on obs it should be using the tag+something | 00:32 |
| mal | Mister_Magister: PR for what? | 00:32 |
| Mister_Magister | for version bump :D | 00:32 |
| mal | nah | 00:32 |
| Mister_Magister | yeah i realised that it was tag causing it, hence i tagged my stuff 1.0 so that it's picked over jolla's | 00:32 |
| Mister_Magister | mal: …that was a joke, with the version bump | 00:33 |
| mal | Mister_Magister: so you have only built it locally? obs should be using tag not spec version | 00:33 |
| Mister_Magister | mal: yes it is using tag | 00:33 |
| Mister_Magister | I just didn't know that 0.15.0 is using tag, now that I've realised it, i've tagged my changes with 1.0 so that my packages will be picked over packages from jolla repo | 00:34 |
| mal | but even without custom tag it would be newer than official version afaik | 00:34 |
| Mister_Magister | no without tag it would be 0.14.4 | 00:34 |
| Mister_Magister | and jolla is using tag 0.15.0 | 00:34 |
| mal | are you sure? | 00:34 |
| mal | on obs that is | 00:34 |
| Mister_Magister | yes thats why i had this issue | 00:34 |
| Mister_Magister | without tag, the version in .spec is 0.14.4 and that's what OBS will use | 00:35 |
| nightishaman | mal: i found it | 00:35 |
| Mister_Magister | you have to use tag 0.15.0 for obs to set version to 0.15.0 | 00:35 |
| mal | obs should be taking tag always | 00:35 |
| Mister_Magister | mal: yes, unless you don't give it a tag | 00:35 |
| nightishaman | init.rc calls enter_default_ns on line 748, which mounts /linkerconfig/default on /linkerconfig | 00:35 |
| Mister_Magister | if you pull from master, HEAD it won't use tag | 00:35 |
| Mister_Magister | it will use .spec version | 00:36 |
| Mister_Magister | because there's no tag provided | 00:36 |
| mal | Mister_Magister: why would you have a fork of sensorfw without all the tags? | 00:36 |
| mal | that makes no sense at all | 00:36 |
| Mister_Magister | mal: you're missing the point | 00:36 |
| Mister_Magister | it doesn't matter what the fork has | 00:36 |
| mal | you are missing the point | 00:36 |
| Mister_Magister | if in tar_git, you supply branch master, and revision HEAD, it will NOT magically change the version to latest tag | 00:37 |
| Mister_Magister | you have to specify tag in revision for it to change version to the tag version | 00:37 |
| mal | obs does all kinds of magic | 00:37 |
| Mister_Magister | i'm telling you how it works | 00:37 |
| Mister_Magister | because i just had this issue | 00:37 |
| mal | and when you did the test you had all of the tags in the fork? | 00:37 |
| Mister_Magister | okay the fork has no tags, i guess it's github's default behaviour | 00:38 |
| mal | forks always have the tags from when you forked it | 00:38 |
| Mister_Magister | nope | 00:38 |
| Mister_Magister | you want me to prove it? | 00:39 |
| Mister_Magister | it could be because of the "Copy the master branch only" that's selected by default | 00:39 |
| Mister_Magister | https://github.com/Michal-Szczepaniak/sensorfw fresh fork, do you see tags | 00:40 |
| mal | wtf, github broke it then recently | 00:40 |
| Mister_Magister | glad we're on the same page | 00:40 |
| Mister_Magister | also yes you're right obs will pull latest tag | 00:40 |
| Mister_Magister | I was wrong | 00:41 |
| nightishaman | mal: So, I figured out why strace shows a two mounts on linkerconfig. The function MountExtraFilesystems mounts linkerconfig as tmpfs again, then later enter_default_ns mounts /linkerconfig/default as /linkerconfig | 00:41 |
| Mister_Magister | mal: also yes, i was correct, it's the default option "Copy the master branch only" that removes tags | 00:42 |
| Mister_Magister | you have to do full fork for tags | 00:42 |
| nightishaman | Mister_Magister: thats why i only do full forks | 00:42 |
| mal | Mister_Magister: so stupid to not have tags of master branch in there, that makes no sense | 00:42 |
| Mister_Magister | well that's microsoft for you | 00:42 |
| Mister_Magister | they haven't made single reasonable decision in years now | 00:43 |
| mal | Mister_Magister: why wouldn't you trust me earlier when I said how obs works, I have used it so much that I know what it does and doesn't do | 00:43 |
| Mister_Magister | yeah you're right, my bad | 00:43 |
| Mister_Magister | sorry | 00:43 |
| mal | nightishaman: is MountExtraFilesystems always run? | 00:45 |
| nightishaman | mal: yes | 00:46 |
| mal | nightishaman: so which call breaks it? | 00:47 |
| nightishaman | the builtin enter_default_ns | 00:47 |
| nightishaman | it gets called on post-fs-data by init.rc and causes /linkerconfig/default to be mounted on top of /linkerconfig | 00:47 |
| mal | so the linkerconfig/default is not ok? | 00:47 |
| nightishaman | yeah | 00:47 |
| nightishaman | but we might want to disable MountExtraFilesystems too because mounting twice is redundant | 00:48 |
| nightishaman | I can make a patch ready and start a PR | 00:48 |
| mal | just wondering why my a16 device doesn't have that issue | 00:49 |
| nightishaman | No idea, but the source code confirms it | 00:49 |
| T42 | <b100dian> mal: is pulseaudio at 17.0 in latest ? | 00:51 |
| T42 | <b100dian> (sorry I don't know who else to ping) | 00:51 |
| mal | @b100dian pa 17 is in sfos 5.1, not in 5.0 | 00:52 |
| mal | on obs latest is still 5.0 | 00:53 |
| T42 | <b100dian> that answers my question. So I need sailfishos:5.1 on devel | 00:53 |
| mal | nightishaman: on my device if I unmount /linkerconfig then /linkerconfig/default has ok content | 00:54 |
| nightishaman | mal: How can I test that on mine? | 00:55 |
| mal | I just did umount /linkerconfig and then checked what is under /linkerconfig | 00:55 |
| Mister_Magister | mal: sensorfwd has -c=/etc/sensorfw/primaryuse.conf does that mean i have to override the primaryuse.conf in sparse? it won't load other configs? | 00:56 |
| nightishaman | ok im gonna check by unmasking droid-hal-init and expand my test script | 00:56 |
| mal | nightishaman: what happens if you just comment out that enter_default_mount_ns from /usr/libexec/droid-hybris/system/etc/init/hw/init.rc | 00:57 |
| nightishaman | if i do umount then i have ok content in /linkerconfig | 00:58 |
| nightishaman | but not in /linkerconfig/default | 00:58 |
| nightishaman | does your device somehow copy stuff to /linkerconfig/default? | 00:58 |
| mal | maybe you have multiple mount layers, umount second time | 00:59 |
| mal | but maybe test that commenting first in init.rc to see what happens | 01:00 |
| mal | Mister_Magister: droid-hal-configs submodule has the default primaryuse.conf, you can override in device config sparse | 01:02 |
| Mister_Magister | well that wasn't my question but that's what I did | 01:02 |
| nightishaman | commenting out enter_default_mount_ns causes it to keep the ld.config.txt | 01:02 |
| Mister_Magister | i was just asking if it can load other configs in that directory so that I don't have to override it | 01:03 |
| mal | nightishaman: are android side processes starting better now? | 01:03 |
| Mister_Magister | who the heck is hogging all the resources with wordle app xdd | 01:03 |
| nightishaman | mal: still having this: 05-31 01:05:17.547 6042 6042 F linker : CANNOT LINK EXECUTABLE "/system/bin/odsign": library "libstatspull.so" not found: needed by main executable | 01:05 |
| nightishaman | and this: droid-hal-init: Control message: Could not find 'aidl/android.frameworks.stats.IStats/default' for ctl.interface_start from pid: 2504 (/system/bin/servicemanager) | 01:05 |
| nightishaman | libstatspull.so is only found in an apex | 01:07 |
| mal | those don't seem very critical processes | 01:08 |
| mal | probably | 01:08 |
| nightishaman | mal: well according to google they are important for ART | 01:10 |
| mal | Mister_Magister: there is a very odd design in sensorfw https://github.com/sailfishos/sensorfw/blob/b62d0e591d736c904128fcfc8476f22cfdbe53b9/core/config.cpp#L63 | 01:10 |
| mal | nightishaman: so not for sfos | 01:10 |
| nightishaman | init.rc waits for it tho | 01:10 |
| mal | hmm where? | 01:11 |
| nightishaman | line 1109 | 01:11 |
| mal | probably different lines in my file | 01:12 |
| nightishaman | then do grep -rn "odsign" | 01:12 |
| mal | comment it out also | 01:14 |
| mal | I assume you mean wait_for_prop odsign.key.done 1 | 01:15 |
| nightishaman | yes | 01:15 |
| nightishaman | and odsign.verification.done in zygote start | 01:15 |
| mal | nightishaman: no | 01:15 |
| nightishaman | what | 01:15 |
| mal | zygote is not started in sfos | 01:15 |
| nightishaman | well i commented it out anyway for testing | 01:16 |
| mal | I meant the other thing | 01:16 |
| nightishaman | yeah i commented out start odsign and odsign.key.done | 01:16 |
| mal | in "on post-fs-data" | 01:16 |
| nightishaman | but if on zygote-start is never run then statsd wont be run and it will keep filling my log with IStats not found | 01:17 |
| Mister_Magister | mal: I seem to be running into race condition where the gesture sensors are not yet available when lipstick starts… do you have better idea than StartPre: sleep 1000000000000 | 01:18 |
| mal | nightishaman: unless something is wrong zygote should not be starting | 01:18 |
| Mister_Magister | I… could loop the app and wait check if plugin is available every second | 01:18 |
| mal | how are the sensors not available? | 01:19 |
| nightishaman | mal: How do I stop these from getting spammed every second then? droid-hal-init: Control message: Could not find 'aidl/android.frameworks.stats.IStats/default' for ctl.interface_start from pid: 2504 (/system/bin/servicemanager) | 01:19 |
| nightishaman | since statsd seems to be started in on zygote-start | 01:19 |
| Mister_Magister | they just aren't sensorfwd says HYBRIS CTL invalid sensor type: 65540 | 01:19 |
| mal | do you see anything about zygote in logs then? | 01:19 |
| Mister_Magister | and even if i restart my daemon sensorfwd won't give it the sensor anymore | 01:20 |
| Mister_Magister | but if i restart sensorfwd it should work fine | 01:20 |
| mal | so sensorfw is started too early | 01:20 |
| Mister_Magister | i would think so yes | 01:20 |
| Mister_Magister | i wonder if there's something i can condition it on instead of just adding random delay | 01:21 |
| mal | maybe some android service activates those? | 01:22 |
| T42 | <b100dian> Doesn't work (re @b100dian: that answers my ques...) | 01:25 |
| T42 | <b100dian> Anyone built 5.1 on obs / devel | 01:26 |
| mal | which device? | 01:26 |
| T42 | <b100dian> https://build.sailfishos.org/project/show/nemo:devel:hw:xiaomi:tucana | 01:26 |
| T42 | <b100dian> But if I had an example of another device on 5.1 with devel I would probably figure it out | 01:27 |
| mal | how about now | 01:27 |
| T42 | <b100dian> hmm :) what did you change mal? Teach the man to fish | 01:28 |
| mal | the order of repos is important in meta | 01:28 |
| mal | https://build.sailfishos.org/project/meta/nemo:devel:hw:xiaomi:tucana | 01:28 |
| T42 | <b100dian> oh, didn't expect that | 01:29 |
| T42 | <b100dian> thanks! | 01:29 |
| mal | the nemo:devel:hw:common is still 5.0 so you had it at the bottom which meant it was used for base stuff | 01:29 |
| mal | which of course is not ok | 01:30 |
| nightishaman | mal: https://pastebin.com/6n2s790f new stuff crashing | 01:30 |
| mal | usb service is useless | 01:30 |
| mal | can be disabled | 01:31 |
| mal | example of a service being disabled https://github.com/mlehtima/droid-config-fp4/blob/master/sparse/usr/libexec/droid-hybris/system/etc/init/disabled_services_device.rc | 01:32 |
| mal | you have to check the .rc files for the service name etc to use | 01:32 |
| nightishaman | yeah but the others are part of the camera | 01:33 |
| nightishaman | all the so files they search are in apexes | 01:33 |
| mal | but now linkerconfig should be ok? | 01:34 |
| mal | so why are those still failing | 01:34 |
| nightishaman | well linkerconfig doesnt have those pathes in it | 01:34 |
| mal | unless there is something wrong in linkerconfig still | 01:35 |
| mal | how different is the linkerconfig if you create it using the linkerconfig in /apex, I remember there might be one there? | 01:36 |
| mal | Mister_Magister: that sensorfw PR is a bit of a mess, why 7 commits? | 01:37 |
| Mister_Magister | why not? i made commits as I went | 01:37 |
| Mister_Magister | you can always squash it | 01:37 |
| mal | 3 commits with same commit message? | 01:37 |
| Mister_Magister | I'm lazy xd | 01:38 |
| mal | some are removing things you added in another commit, that's not ok | 01:38 |
| Mister_Magister | you can always squash it xd but if it is a problem i'll squash it myself | 01:39 |
| nightishaman | mal: when i run linkerconfig right now after starting droid-hal-init, it gives a more expanded linkerconfig which also include libstatspull.so | 01:39 |
| nightishaman | however the camera hal is still not included in that linkerconfig | 01:40 |
| Mister_Magister | here you go, squashed | 01:40 |
| nightishaman | using the one in /apex/com.android.runtime makes no difference | 01:40 |
| mal | ok, something odd still with that | 01:40 |
| T42 | <b100dian> (yee I have sound back on jb2q) | 01:40 |
| Mister_Magister | I've added delay 100 to sensorfw start xd that just made it not start, but 10s was not enough :/ | 01:44 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!