<MarioMariollini>
Hi. I am struggling to generate an sdk for powerpc using multilib. I have followed section 17.2 from the docs and added the require and default tune to my multiconfig, like so:
<MarioMariollini>
The system is running on ppc 64-bit but I need to run a third party binary that is 32-bit. However, when I try to run the do_populate_sdk task I run into an error with do_generate_content trying to touch a file in a directory that does not exist:
<MarioMariollini>
mypowerpc64 is the name of the machine and multiconfig. the directory that does not exist is powerpc under opt/sdk, it is only powerpc64
<MarioMariollini>
Any hints on how I should proceed to debug or fix this?
<MarioMariollini>
sorry, failed to copy the directory: /home/mario/build_dir/tmp-glibc/work/x86_64-nativesdk-oesdk-linux/meta-environment-mypowerpc64/1.0-r8/sdk/image//opt/sdk/powerpc/4.3.0/site-config-powerpc-oemllib32-linux
khazakar has joined #yocto
<RP>
MarioMariollini: powerpc is one of the lesser used architectures now and therefore isn't as widely tested/used. You're the first person in a long time I've seen trying to use multilib with it
<RP>
MarioMariollini: that sounds like it is the code trying to gather site files., You could disable that bit and see if that is the only issue
Guest55 has joined #yocto
Guest55 has quit [Client Quit]
vladest has quit [Remote host closed the connection]
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
<MarioMariollini>
RP: added a mkdir -p to create that missing dir in meta-environment.bb:create_sdk_files instead. Now the error comes in gdb-cross-canadian trying to find python... seems like it will be an uphill battle
<rburton>
RP: can you merge the qt/webos repo change to yocto-ab2?
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
alessioigor has quit [Remote host closed the connection]
alessioigor has joined #yocto
neofutur has quit [Ping timeout: 246 seconds]
vladest has joined #yocto
neofutur_ has joined #yocto
zpfvo has quit [Quit: Leaving.]
khazakar has quit [Quit: Connection closed for inactivity]
khazakar has joined #yocto
neofutur_ has quit [Ping timeout: 246 seconds]
neofutur_ has joined #yocto
samkent has quit [Ping timeout: 264 seconds]
florian has joined #yocto
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
<khem>
How can we get different group id's when chgrp is used in do_install in a recipe which is also inheriting useradd, the ipk had one 999 for ptest and packages using the same gid, however, in the final image gid for ptest user changes to 995. I can imagine that at image build time, image builder collects all needed uid/gid's as it can see all of them whereas when building just one package it does not see users being created by other
<khem>
recipes and perhaps that ordering changes it ?
<khem>
image had 995 for ptest gid
<khem>
is there some logic is uses to assign gid's during image creation ?
<khem>
how is ./qemux86_64-yoe-linux-musl/core-image-ptest-cpio/1.0/rootfs/etc/group populated ?
<khem>
this is same as what I posted in pastebin
<fray>
starts with the one your distro provides, then the adduser(group) should populate it during package installs
<khem>
hmm so does it mean some package is creating a group and has 995 for that group's gid ?
<khem>
how does it get gid value
<fray>
so if your distro one has the group already it will use that (static defined), otherwise it will use the one dynamically added by the addgroup, which could be a static group or dynamic (next available) so package install ordering.. in the temp directory lots of groups can get re-used during building in that case
<fray>
If two packages both need a dynamic gid, they BOTH need to add it
<khem>
I am not using static ids so that part is out of picture
<fray>
(package install order is not guarenteed, thus each package using a dynamic gid needs to add it. You can use some tricks to enforce some ordering, but it's not a good idea)
<khem>
so you mean each one needs to inherit useradd class and dilligently add all the groups/users it needs ?
<khem>
right
<fray>
(not a good idea = different package managers order things differently.. so you could enforce the order for say RPM, but deb might not work..)
<fray>
yes
<khem>
I am using ipk
<fray>
I just know rpm and deb order things slightly differently.. I assume ipk has it's own quarks..
<khem>
during do_install it creates gid=999 for ptest and during image creation ptes gid becomes 995
<fray>
NOTE: ptest-runner: user root already exists, not re-creating it
<fray>
Running useradd commands...
<fray>
that says something tried to create the group 'ptest' with a root user, I think
<khem>
if you look into ptest-runner recipe then it seems to not do that
<fray>
ok.. -g is a group id to set the new user to... -G is additional supplmentary groups ok..
<fray>
so pteat will be in the group 'ptest' and added to the group 'root'
<khem>
right thats what I thought
<khem>
problem is next useradd is from cpio and by then user/grp ptest is already created but has different value for gid
<fray>
something is odd though.. the error is user root already exists.. so it's reading that last parameter (after -G) as the user to create, I think
<khem>
yeah maybe last param if not read is treated as 0 which is root
<fray>
I'm suspicious there is a problem somewhere in the arguments of that 'useradd'. Normally when I do this though I break it into two steps.. add the group and then I add the user.. (it can be done as a single operation, but I tend not to)
<fray>
Ya, I'm really not sure here.. something strange is happening though
<khem>
-g ptest -G root from where does it add -G root I wonder
simonew has joined #yocto
<fray>
ya, that is what is confusing me.. something doing a :append somewhere?
xmn has joined #yocto
florian has quit [Ping timeout: 256 seconds]
kpo_ has joined #yocto
Haxxa has quit [Quit: Haxxa flies away.]
MarioMariollini has quit [Quit: Client closed]
Haxxa has joined #yocto
vladest has quit [Remote host closed the connection]
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
vladest has joined #yocto
simonew has quit [Quit: Konversation terminated!]
florian has joined #yocto
sudip_ is now known as sudip
prabhakarlad has quit [Quit: Client closed]
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
alessioigor has quit [Quit: alessioigor]
alessioigor has joined #yocto
prabhakarlad has joined #yocto
prabhakarlad has quit [Ping timeout: 250 seconds]
mvlad has quit [Remote host closed the connection]
ederibaucourt has quit [Quit: ZNC 1.8.2 - https://znc.in]
ederibaucourt has joined #yocto
ykrons has quit [Ping timeout: 260 seconds]
ykrons has joined #yocto
<khem>
fray: I removed ptest-runner from image for checks and now the user ptest is only created by single package which is cpio-ptest and the problem still happens, gid in ipk does not match the gid in image
<khem>
this is going to be an issue for adopting ptests widely where they must be run under non-root user
xmn has joined #yocto
<RP>
rburton: done
<khem>
abelloni: you need to combine the musl systemd patch with systemd 255 bump together, I see that musl refresh patch is applied in your master-next but the patch to upgrade systemd to 255 is not