<Guest28>
Hey, Iosing my mind a little over "do_populate_lic: Could not copy license file" issue
<Guest28>
Im trying to bring over the License from a github repo..
<Guest28>
Happy to talk more if anyones around to help
<Guest28>
Never mind guys - think I got it
Guest28 has quit [Quit: Connection closed]
ncaidin_lf has joined #yocto
ncaidin_lf has quit [Client Quit]
wing0 has joined #yocto
wing0 has left #yocto [#yocto]
kpo_ has quit [Read error: Connection reset by peer]
wing0 has joined #yocto
vmeson has joined #yocto
kpo_ has joined #yocto
<khem>
yates: yes not from same repo but yes technically from its upstream gcc sources
<jordemort>
derRichard: I seem to have figured it out, looking at the dmesg, when I pop a shell in my initramfs, the `imx-busfreq` driver kicks in and scales the CPU frequency down, and then it doesn't seem like the NAND works anymore; if I'm quick and disable the bus frequency scaling before it kicks in, it works fine. If I let the machine boot all the way it mounts the NAND just before the bus frequency scaling seems to kick in and then it never kicks in
<jordemort>
and everything is fine
<jordemort>
(this is an imx7 sabre sd derivative)
otavio has quit [Remote host closed the connection]
otavio has joined #yocto
rcw has quit [Quit: Leaving]
behanw has quit [Quit: Connection closed for inactivity]
<yates>
obviously the problem is that there isn't one in the qemucsky-poky-linux work folder. why not?
<alex88>
is there a way to check for overrides in do_install? so I can run a command for an image but not for another?
<alex88>
It's my own recipe, should I just use a bbappend?
mihai has quit [Ping timeout: 244 seconds]
<yates>
alex88: i would think so. my guess (to be taken with a grain of salt) is to a) find the do_install (e.g., in a .bbclass), and b) add code to do the check for the image in a preamble of the do_install.
<yates>
it is "illegal" to modify .bbclass files in poky, but i usually do it as a starting point, then back it out somehow into my own layer later.
<yates>
khem: are you still up?
<yates>
anyone have any thoughts on the "missing libgcc.a" problem?
Vonter has joined #yocto
paulg has quit [Ping timeout: 268 seconds]
camus1 has joined #yocto
camus has quit [Read error: Connection reset by peer]
camus1 is now known as camus
<alex88>
yates, oh ok thanks, well the bbclass is my own, nothing that comes from any official layer, thanks for the advice!
elfenix has quit [Ping timeout: 252 seconds]
<alex88>
if I do SETTINGS_FILENAME = "settings-secure.js" in my recipe, how can I get that variable availble in do_install?
camus has quit [Read error: Connection reset by peer]
camus1 has joined #yocto
camus1 is now known as camus
<alex88>
nvm it won't work anyway as I can't set the override based on the iamge name
rob_w has joined #yocto
zyga-mbp has joined #yocto
<RP>
sgw: I think the autobuiler is hanging on that smp change. It does show in the logs though!
LocutusOfBorg has quit [Remote host closed the connection]
LocutusOfBorg has joined #yocto
kayterina has joined #yocto
kayterina_ has joined #yocto
kayterina_ has quit [Client Quit]
LocutusOfBorg has quit [Remote host closed the connection]
kpo_ has quit [Read error: Connection reset by peer]
LocutusOfBorg has joined #yocto
kpo_ has joined #yocto
prabhakarlad has quit [Quit: Client closed]
xmn has quit [Quit: ZZZzzz…]
prabhakarlad has joined #yocto
TommyGilchrist has joined #yocto
TommyGilchrist is now known as tommyg
<fabatera[m]>
Hi! In case it helps somebody, feed back from my own question from some days ago: host is NUC10, Using poky and meta-intel (Zeus) and trying to nfsroot, loading the e1000e Intel driver (during boot or initramfs shell) is not creating the ETH interface during boot, so nfsroot fails. Booting same image from pen drive, network is OK.
<fabatera[m]>
After I changed to Gatesgarth, it worked almost out of the box. Only need to include nfsroot to initramfs image.
Guest5 has joined #yocto
<Guest5>
Hi all, I'm trying Yocto with WSL2/WSLg, everything look okay except the devveloper shell, new terminal is open but i'm drop inside the user home directory and not inside the recipe devshell folder. any idea to fix that ? thx.
elfenix has joined #yocto
Guest530 has joined #yocto
<Guest530>
same issue with devpyshell call :S
Guest5 has quit [Ping timeout: 250 seconds]
goliath has joined #yocto
Guest530 has quit [Ping timeout: 250 seconds]
prabhakarlad has quit [Quit: Client closed]
<derRichard>
my distro sets NO_RECOMMENDATIONS to install only what is really needed. but gcc-sanitizers.inc does RRECOMMENDS_${PN} += "libasan libubsan"
<derRichard>
how can i get libasan installed on my target without unsetting NO_RECOMMENDATIONS?
tommyg has quit [Ping timeout: 250 seconds]
<derRichard>
or is it less complicated than i assumed? just statung libasan as dependency? hmmm
<LetoThe2nd>
derRichard: i kind of shied back from suggesting to do a RDEPENDS_pn-whatever += "libasan"
<LetoThe2nd>
for the distro, that is.
<derRichard>
why that?
<LetoThe2nd>
because its ugly?
<derRichard>
what is the less ugly way? :D
<LetoThe2nd>
you could obviously put it right next to the NO_RECOMMENDATIONS and add a note why its there, making it still ugly but understandable
<LetoThe2nd>
i don't know of any pretty way, sorry.
<derRichard>
ah you mean that...
<derRichard>
doing embedded systems is always ugly ;)
<LetoThe2nd>
maybe not exactly always, but a substantial number of times, yes.
<derRichard>
non-trivial embedded systems ;)
<derRichard>
packing the sdk is also fun here
<derRichard>
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
<Tyaku>
Hi, I am testing Yocto on a Raspberry PI and at terms the idea is to run Yocto in IMX8. But I have a question about memory requirements. because I don't see any requirements in Yocto website
<Tyaku>
Did we know the minimal size of FLASH and RAM required to run Yocto ?
<Tyaku>
On raspberry PI the "image" (that is flashed on SD card) is about 653Mo!
<rburton>
a core-image-base is actually about 50mb or so
<rburton>
images can have lots of empty space
<rburton>
and obviously, depends what you put in it
<rburton>
an image that boot straight from kernel to app will be tiny and have minimal memory requirements
<Tyaku>
I build an image using core-image-base for raspberry pi and 'xxxxx-raspberrypi3-20210617113831.rootfs.wic.bz2' takes 135 Mo, unziped the wic file is 650Mo
<LetoThe2nd>
Tyaku: core-image-minimal on the rpi is in the range of single digit MBs
<Tyaku>
How is it possible ?
prabhakarlad has joined #yocto
<LetoThe2nd>
Tyaku: possible because you added random stuff?
<LetoThe2nd>
Tyaku: look at the packages and manifests that make up the image
<Tyaku>
hum, not so much, I was doing a getting started tutorial, the unique stuffs I added is "nano" and "ntpd"
<Tyaku>
Hum
pbergin has joined #yocto
<dwagenk>
Tyaku: using core-image-base adds a lot of stuff you will probably not need in your own images. If you want to know about minimum size build core-image-minimal instead, as Leto suggested.
<LetoThe2nd>
Tyaku: which "getting started tutorial" would that be?
<Tyaku>
I realize That i also have "splash tools-debug tools-profile tools-sdk" in my image bb.
<LetoThe2nd>
the tools-* are huge. so there you go.
<Tyaku>
Yeah and I have to find how i get this, because i made my .bb file based on core-image-base.bb
<Tyaku>
But no way, I'm going to try with core-image-minimal
<rburton>
Tyaku: the wic will have empty space in, which you're counting in the image size
<rburton>
boot it and look at disk usage (or look at how big the rootfs.tar.gz is, not the final wic)
<rburton>
i had a 2gb wic image last week. gzip took it down to 50mb. lots of empty space :)
<LetoThe2nd>
space... the final frontier...
<Tyaku>
Currently, my rootfs.tar.bz2 is 34Mb
<LetoThe2nd>
can anybody tell me why my brain just realized that what the bitbake community *REALLY* needs is a tool named "gyros"?
<rburton>
it does?
<Tyaku>
Ok, but now what about the required memory in the Flash for the entire system, if the rootfs is about 50Mb, i also have to count the space for bootloader isn't it ?
<LetoThe2nd>
rburton: it does. i know it.
<rburton>
the bootloader is tiny
<Tyaku>
bootloader/kernel/rootfs
<LetoThe2nd>
Tyaku: it depends so much on your payload and requirements. people are building distros in 10M, people are building distros in 100M, people are building distros in 1G
Guest5 has joined #yocto
<Guest5>
Hi all, I'm trying Yocto with WSL2/WSLg, everything look okay except the devveloper shell, new terminal is open but i'm drop inside the user home directory and not inside the recipe devshell folder. any idea to fix that ? thx.
<rburton>
Guest5: filing a bug might be a good start
<Tyaku>
Currently we don't really know our requirements thats why I try to check "with the base image"
<LetoThe2nd>
Guest5: maybe ask on the ML
<LetoThe2nd>
Tyaku: but then there can be no answer.
<Tyaku>
How to know the size of the bootloader ? Is it build with bitbake and included in the image ?
<Tyaku>
What about the kernel ?
<LetoThe2nd>
Tyaku: the bootloader is counted in kBytes. forget about it.
<Tyaku>
What about the kernel ? Is it included in rootfs or another file ?
<LetoThe2nd>
Tyaku: try to understand. if you ask for the requirements now, and we say: 32M ram, 4M flash. then thats technically totally legit, because you *CAN* build distributions in that size. however, you don't learn anything because it in no way tells you anything about the requirements of YOUR project.
<LetoThe2nd>
Tyaku: the kernel can be in the rootfs, it can be separate. however it will probably be in the wic at least in the vast majority of cases.
<LetoThe2nd>
Tyaku: so it is really, really not about "yocto requirements". it is about "your requirements."
<LetoThe2nd>
for example: do you need systemd? do you need graphics? do you need opengl? do you need qt? and so on, and so on.
<Tyaku>
Nop nothing of this, The device will act as a gateway (thread/zigbeee).
<Tyaku>
As our code is not on linux currently, I don't have an idea for the requirements
<LetoThe2nd>
then you will probably end up in the below 128M of flash range at least. no idea about the ram requirements of your code
<rburton>
a well tuned kernel and tiny userspace to be a gateway can be nice and small
<LetoThe2nd>
*can*
<LetoThe2nd>
then some c-level comes along and says "we need this fancy web ui"
<LetoThe2nd>
so again, its really about "the requirements"
<kayterina[m]>
a tool gyros-souvlaki? 'bitbake gyros' sounds delicious
mckoan is now known as mckoan|away
<LetoThe2nd>
bitbake -c gyros core-image-minimal
<LetoThe2nd>
:)
<rburton>
yeah adding a nodejs web ui makes it not-tiny by definition
<LetoThe2nd>
rburton: hire me to prove otherwise :)
<rburton>
haha
pbergin has quit [Ping timeout: 264 seconds]
<rburton>
I wonder why gitlab thinks meta-arm is 10% PHP code
leon-anavi has quit [Remote host closed the connection]
leon-anavi has joined #yocto
kpo_ has quit [Read error: Connection reset by peer]
<Guest2849>
Hey everyone trying to get some help with "LIC_FILES_CHKSUM points to an invalid file:" I keep getting
<LetoThe2nd>
Guest2849: well then fix the variable? look at it, and make sure it is set to a value that exists in the sources of the package
<Guest2849>
Thats the thing, The variable and all looks good. Setting it to pull from the repo.. just having a hard time figuring out what it doesnt like...
<LetoThe2nd>
given that desciption its hard to say anything other than "look again, triple check everything"
<Guest2849>
cool, Ill give it another shot
<LetoThe2nd>
Guest2849: maybe just put the recipe into a pastebin
<Guest2849>
heres the minimal recipe Ive been using to test out the LIC_FILES_CHKSUM with..
<Guest2849>
prolly just something dumb im doing, not quite sure what though
<LetoThe2nd>
Guest2849: S = "${WORKDIR}/git", probably
<Guest2849>
appreciate it LetoThe2nd
<Guest2849>
yep that sure does make things a lot better
<Guest2849>
let me try to wrap my head around the S variable ..
xmn has joined #yocto
<Guest2849>
so what is the default for S? LetoThe2nd
paulg has joined #yocto
<rburton>
see bitbake.conf. WORKDIR/BPN-PV, so if your recipe is foo_1.0.bb the default is foo-1.0, matching the naming of most software tarballs
<Guest2849>
Im trying to understand why LIC_FILES_CHKSUM can only get the the LICENSE file when S="${WORKDIR}/git and not the default value..
leon-anavi has quit [Remote host closed the connection]
leon-anavi has joined #yocto
<perdmann_>
Hi, i need to know how i can use an azure PAT (Access token) in a Bitbake recipe :)
gsalazar has joined #yocto
<Guest2849>
Im also trying to figure out how I can run the python bdist_wheel command and pip install from within a recipe. Mostly what all should I inherit etc
<perdmann_>
Guest2849: inherit setuptools3 e.g.
gsalazar has quit [Remote host closed the connection]
<Guest2849>
Alright - im kinda new at this so what layer should I be adding for setuptools3 ?
rob_w has quit [Quit: Leaving]
gsalazar has joined #yocto
<Guest2849>
perdmann_: thanks for chiming in - appreciate it. Thin I got a long way to go with this recipe, so really appreciate al the help :)
<yates>
obviously the problem is that there isn't one in the qemucsky-poky-linux work folder. why not?
<gourve_l>
hi! Is there a recipe which version (PV) is automatically incremented/updated when DISTRO_VERSION changes? My goal is to be able to do "opkg update; opkg upgrade" to update a file (like /etc/os-release) without having to change the PV of a recipe manually
<Guest2849>
yates: maybe check the dir you're running you bitbake command from. sometimes the paths are setup a certain way. also check you didnt sudo or soemthing
<perdmann_>
Guest2849: sorry, iam not deep into this stuff, i just know that a friend of mine did a recipe with this and it works
<Guest2849>
no worries perdmann_
<yates>
s/obviously//
<yates>
it is not obvious
camus has quit [Quit: camus]
<Guest2849>
I start seeing the maximum depth exceeded error when I add inherit pypi or setuptools in my recipe. Im testing out the recipe using bitbake -b
gsalazar has quit [Quit: Leaving]
zyga-mbp has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<rburton>
don't use -b
<Guest2849>
oh shoot, what should I be using then?
<rburton>
bitbake <recipename>
<Guest2849>
I see, thanks rburton
<gourve_l>
(forget my question, 'PV="${DISTRO_VERSION}"' seems to work in fact)
Guest2849 has quit [Ping timeout: 268 seconds]
ilunev has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
vmeson has quit [Quit: Konversation terminated!]
zyga-mbp has joined #yocto
vmeson has joined #yocto
Guest28 has joined #yocto
zyga-mbp has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
prabhakarlad has quit [Quit: Client closed]
prabhakarlad has joined #yocto
<Tyaku>
I have another question, it's not "directly" related to poky but it's possible that you had an idea.. I try to create a Yocto environement for IMX8, using this guide: https://www.nxp.com/docs/en/user-guide/IMX_YOCTO_PROJECT_USERS_GUIDE.pdf When I build Yocto, at some point i have an error:
<Tyaku>
ERROR: Task (/home/user/imx-yocto-bsp/sources/meta-imx/meta-bsp/recipes-kernel/linux-libc-headers/linux-libc-headers_5.10.bb:do_install) failed with exit code '134'
<rburton>
you have the file at meta-opentrons/recipes-example/open_git.bb
<rburton>
but the classic glob is recipes-*/*/*.bb
<rburton>
so you either need to rewrite the glob to include recipes-*/*.bb, or move the recipe into a subdirectory
<Guest28>
where is the classic glob declared at?
<rburton>
in the layer.conf
<rburton>
inside that layer
<Guest28>
i see, ok
<Guest28>
Thanks a lot rburton, let me try that
<Guest28>
really appreciate the help.
<Guest28>
Im trying to keep the recipe from doing anything on its own. Thin right now looking at some make files etc in the repo it's trying to run do_package_write_ipk_setscene: on its own. What function can I overwrite to keep that from happening?
<Guest28>
think*
camus has joined #yocto
<rburton>
why would you do that?
<rburton>
you're trying to disable the recipe from working
<rburton>
unless you mean the recipe has a makefile that explicits calls do_package_write_ipk_setscene by itself which is terrifying
<Guest28>
the repo has a make file .. which is making he recipe explicitly call do_package_write_ipk_setscene by itself ..
<Guest28>
at thats how I understand it
<Guest28>
I can pastebin some more logs if that helps
<rburton>
all recipes inherit package.bbclass which will end up calling do_package_write_ipkg_setscene because that is how packaging works
<rburton>
ok so something in your setup is fundamentally broken
<rburton>
have you put deploy/ outside of tmp/?
kayterina has quit [Remote host closed the connection]
<rburton>
if so, delete deploy/ too
<rburton>
normally it's under tmp/ so you can just delete everything that is built in one go
<Guest28>
yeah i got a deploy out side of tmp
<rburton>
right so the warning is 100% accurate. there are files in deploy it doesn't know about, because you just deleted the index of files it knows about
<rburton>
i recommend putting deploy in tmp so you can just wipe everything that can be recompiled with a single rm. or next time, remember to delete both.
<Guest28>
got it.. made of note of that.. thanks
camus has quit [Read error: Connection reset by peer]
camus1 has joined #yocto
<Guest28>
oh that looks beautiful - my open recipe isnt spitting out those errors now
<Guest28>
thanks a lot. would been chasing my tail forever on this one
Net147 has quit [Remote host closed the connection]
<Guest28>
I'll get started with the real task now..
<Guest28>
Mind if I run that by you, just to see if you think there's anything fundamentally wrong with the thought?
Net147 has joined #yocto
Net147 has joined #yocto
Net147 has quit [Changing host]
camus has joined #yocto
camus1 has quit [Ping timeout: 264 seconds]
<rburton>
what thought?
<Guest28>
so I got a repo with a bunch on python source. With in the recipe I want to make a .whl file, and then install that on the host using pip. All from within the recipe. Does that thought sound right somewhat?
<Guest28>
Trying to use the bdist_wheel command to make the whl file. So was thinking the setuptools class can help me with it in the recipe?
<rburton>
you're overcomplicating things
<rburton>
write a setup.py to install the python like usual
<rburton>
inherit setuptools3
<rburton>
done
<rburton>
setuptools3 will install the python as a standard third party library, without faffing about using wheels and pip. you're building a distro, those tools are for users.
<Guest28>
got it. Thanks rburton. I'll give it a whirl.
LetoThe2nd has quit [Quit: Connection closed for inactivity]
<yates>
rburton: who is the expert in this area?
<yates>
in a nutshell, i'm not sure if a) the libgcc.a is not getting generated into the correct folder, or b) the flags are pointing ld.bfd to the wrong folder
<yates>
b) the flags (or lack thereof)...
<perdmann_>
yates: sorry iam late to the party. how does your makefile look?
<yates>
perdmann_: thanks - hang on and let me get at that.
<perdmann_>
yates: in this one log, your calling make with HOSTCC=cc so that looks like you are not using cross compiler on this
<rburton>
yates: well i'd compare the build to eg qemuarm which is obviously known to work
<rburton>
HOSTCC should be the host compiler, not the cross compiler, thus the name
<rburton>
the error is from the cross ld
<yates>
rburton: yes, that much is clear
<perdmann_>
ok i see, so you try to compile the linux kernel with the SDK ?
<perdmann_>
did you already look on the linking paths your linker has?
<yates>
i'm not even sure there is a makefile involved. this do.run_compile script and make invocation are so complex it is difficult for me to determine.
<yates>
it could be make taking all the environment variable and command line inputs and doing the builds itself by its own internal rules
<yates>
in short, this is beyond my make-fu..
<yates>
but it looks to me that the paths are passed in the CC command with several "-L" options, but then the linking is done outside of the compilation with the ld.bfd, and no path flags are passed to it.
<yates>
that's my best guess with my limited understanding.
<yates>
perdmann_: you did see the run.do_compile and log.do_compile pastebins i submitted above, right?
wing0 has quit [Quit: WeeChat 3.1]
<yates>
LD="csky-poky-linux-ld.bfd "
<yates>
again, at the end-of-the-day, the error i'm chasing is this: csky-poky-linux-ld.bfd: cannot find libgcc.a: No such file or directory
jpuhlman has joined #yocto
<rburton>
honestly, compare a build for qemuarm
<rburton>
is this linux-yocto or your own kernel?
<yates>
the mainline kernel
<yates>
not linux-yoct
<yates>
not linux-yocto
<rburton>
build that kernel recipe for qemuarm and see if it works
<rburton>
if it does, then its your toolchain
<rburton>
if it fails, its your kernel
<yates>
rburton: so you're saying edit my build/conf/local.conf and set MACHINE ??="qemuarm"?
<rburton>
yes
<yates>
ok that's easy, i can try that
<perdmann_>
Oh so you build with bitbake, thats weird.
<perdmann_>
yates: you can look at your-ld --verbose | grep SEARCH_PATH
<perdmann_>
yates: you can look at your-ld --verbose | grep SEARCH_DIR
<perdmann_>
Forget the first one
<yates>
perdmann_: wha? since when is building in yocto using bitbake weird?
<perdmann_>
yates: weird is that this is not working :-D
<yates>
i'm rerunning the building using MACHINE ??= "qemuarm" as rburton suggested right now.
<yates>
s/building/build/
<yates>
that is a good idea and i will try it in a few
zyga-mbp has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<khem>
yates: some kernel ports need libgcc seems yours is one of those so please add DEPENDS_append = " libgcc" in kernel recipe or bbappend and that perhaps will help
Guest28 has quit [Quit: Connection closed]
* derRichard
wonders why we can't run opkg in parallel while installing packages into a rootfs. opkg uses almost all cpu time in the decompressor.
<derRichard>
did someone of you look into this? should be doable with minimal changes to opkg
vmeson has joined #yocto
LetoThe2nd has joined #yocto
Guest28 has joined #yocto
<Guest28>
whats a good way of finding out if the oe-core i got has a recipe for a particualr python module, e.g. fastapi
<Guest28>
more like if I have a setup.py with an array of install_requires, would setuptools3 take care of them all, or do I have to pull in recipes for them all?
<LetoThe2nd>
all dependencies have to be seperately pacaked.
<LetoThe2nd>
packaged. pancaked. /me tries to untangle fingers, fails.
<Guest28>
got it. Thanks LetoThe2nd
<Guest28>
is there an oe-core recommended way of writing recipes for python modules?
<Guest28>
guess I can just look at an existing one. was wondering if theres a script or something
<Guest28>
or a devtools way of going about it
<LetoThe2nd>
you can obviously try devtool add, but AFAIK most pyhton module recipes are rather trivial - just at a couple of those.
<LetoThe2nd>
leon-anavi: ^^
roussinm has quit [Quit: WeeChat 3.3-dev]
<leon-anavi>
LetoThe2nd, +1 yes, writing recipes for meta-python is straight-forward: just a few lines in most of the cases.
<Guest28>
something like this for istance, I can just make do with the .bb file, I dont have to bring in the entire meta-layer its part of?
<LetoThe2nd>
Guest28: ask leon-anavi, he's the god of python modules
<Guest28>
oh I should be putthing them under meta-python?
<Guest28>
damn, leon-anavi
<Guest28>
you free to talk rn?
<leon-anavi>
hehe, not really, I am doing straight-forward things with meta-python. Nothing fancy.
<Guest28>
Im porting out this python based sourced to this new soc that uses yocto
<leon-anavi>
is it hardware specific?
<Guest28>
yeah its toradex uses some nxp chip
<leon-anavi>
ah, ok
<leon-anavi>
are you missing any python packages in meta-python that your stuff requires as dependencies?
<Guest28>
and the previous system I'm porting from was doing things in some funky way .. like they would make whl files first then run pip on them. I took rburtons advice this morn and am going for a simple setup.py to take care of everything
xmn has quit [Ping timeout: 268 seconds]
<Guest28>
yeah I cant seem to find 'fastapi==0.54.1', in meta-python
<Guest28>
well the bsp Im building on top of doesnt even have meta-python, so I guess I can bring the layer in
<Guest28>
but the layer doesnt seem to have a few python modules I would need
<Guest28>
can I just clone meta-python and add it to bbconf and all
<Guest28>
or do I need to use a bitbake command to bring in existing layers?
<leon-anavi>
well... I think python3-fastapi recipe is always welcome to meta-python.
<Guest28>
I guess I can upstream it
<leon-anavi>
however it looks like the latest version is newer than 0.54.1. Better use latest and greatest stable.
<Guest28>
once I wrap my head around everything
<Guest28>
got it
<leon-anavi>
I didn't understand the question about cloning. Of course you can clone the repo. But what exactly are you asking for?
<Guest28>
in order to take care of the dependencies for my image
<moto_timo[m]>
To see what Oe-specific sub packages are needed
<leon-anavi>
Guest28, you will need the standard stuff: to add meta-python to bblayers.conf (if it is not already there). After that make sure to add DEPENDS and RDEPENDS_${PN} in your recipe for all build and runtime dependencies.
<Guest28>
ok, I see
<Guest28>
and to make sure my everything gets added on top of my current image what am I doing
<Guest28>
some append in local.conf?
<Guest28>
think it was IMAGE_INSTALL_append or something?
zyga-mbp has joined #yocto
<Guest28>
still a little lost .. if I use devtool to generate the recipe for fastapi, what do I give in as the source tree or URI ?
<moto_timo[m]>
See the command I posted above
<moto_timo[m]>
Literally start with that
<Guest28>
oh so just use the file.pythonhosted as the source URI .. ok
zyga-mbp has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<moto_timo[m]>
That is what pypi bbclass uses, so when you drop SRC_URI and inherit pypi it just works
<moto_timo[m]>
The pypi bbclass sets the SRC_URI for you.
<Guest28>
ok, thats real nice. Thanks for walking me through this!
Vonter has quit [Ping timeout: 268 seconds]
<Guest28>
what was thing about oe-specific packages again?
<moto-timo>
we split up python into smaller subpackages. so by default recipetool will RDEPENDS on python3-core, but you might need more than that
<moto-timo>
those other subpackages are defined in the python3-manifest.json I linked to earlier
<Guest28>
so the fastapi I have right now, I should add all those subpacked to RDEPENDS ?
<Guest28>
subpackages
<moto-timo>
just the ones it needs. the whole point of yocto project is to build ONLY what you need and include ONLY what you need
<Guest28>
got it
<moto-timo>
but you need to figure that out for yourself, either from the hints from recipetool or by trial-and-error running on target/qemu
Vonter has joined #yocto
<moto-timo>
but basically I have just described to you exactly the workflow I would have done if I needed python3-fastapi
<Guest28>
got it
<Guest28>
thats super helpful
<moto-timo>
now it's your turn to fly
<moto-timo>
:)
<Guest28>
those sub packages wont need their own recipes or something, would it?
<moto-timo>
a lot of those warnings looked like they were about docs... we usually don't package docs so those might be able to be ignored (unless you want docs on target)
<moto-timo>
no, just e.g. RDEPENDS_${PN} += "python3-numbers" or whatever is needed
<moto-timo>
but if you look on pypi for "numbers" you won't find it , it's in the standard libraries
<moto-timo>
but we do things differently, so you need to understand what is in python3-manifest.json (and the packages-split folder in your built python3)
<Guest28>
let me just quickly read up a little on RDEPENDS_${PN} first and get back to you
<Guest28>
cant remember what these variables are for to save my life. spent the morning reading about S, I still have my doubts
<Guest28>
dare I risk asking another trad newcomer questions, you got any recommendations for an irc client?
<Guest28>
trying to run it on my ec2 instance
<moto-timo>
depends on what you are comfortable with. command line? weechat GUI? I use IRCCloud client... others use other GUI clients
<Guest28>
like a headless client or something I can keep running in screen
<moto-timo>
mosh, tmux, weechat
<moto-timo>
mosh keeps the ssh session alive, tmux keeps the console alive, weechat is a fairly nicely featured text UI for those that like command line interfaces
<Guest28>
cool thanks Tim. And again, im sorry for having you walk me through some of the rather simpler stuff. Guess it can just be a little overwhelming at times.
<moto-timo>
you can also use Element with Matrix
<moto-timo>
we all understand the ramp up... we all had to learn it
<moto-timo>
but we also might get irked when it becomes "let me google that for you" or "let me copy the section of the docs that completely explains your question"
<moto-timo>
depends on how the week is going ;)
<leon-anavi>
bed time here. See you and have fun with meta-python :)
leon-anavi has quit [Quit: Leaving]
<moto-timo>
some stuff like workflow I described earlier isn't really captured anywhere
<Guest28>
I just got rusty havent used yocto/oe-core in a while.. I see stuff from you on the oe-core mailing list a whole lot you and khem.. think I got on it when I upstreamed a recipe for nlohmann json
<Guest28>
Im just trying to get some stuff ported to this toradex soc.. and I spent way too much time dealing with how theyve setup their bsp using the repo tool .. and how their stuff was breaking last week. trying to make up for that a little
<Guest28>
ill get back to it .. Thanks again!!
<moto-timo>
you're welcome
<moto-timo>
and yes, bsps can be.... special
wesm has quit [Ping timeout: 268 seconds]
wesm has joined #yocto
override has quit [Killed (mercury.libera.chat (Nickname regained by services))]
override_ has joined #yocto
Vonter has quit [Ping timeout: 268 seconds]
<yates>
rburton: ok, so qemuarm ran fine.
prabhakarlad has quit [Quit: Client closed]
Gaffel has quit [Quit: What's that?]
Gaffel has joined #yocto
<yates>
but it's evidently using linux kernel 5.8.18 instead of the 5.12 kernel recipe i have in my layer
<Guest28>
moto-timo: not to rant, but I just had the strong urge to put this on the record. In our earlier conversation you made references to manifest file. I havent come across them a whole lot in the past, so I naturally look them up in the yocto manual. Which again just goes around in circles - doesnt really talk about the kind of manifest you pointed me
<Guest28>
to earlier. Not trying to use you as a consulting service, just letting you know the lay of the land is such that one has to ask questions! A lot of questions even..
<moto_timo[m]>
The docs are now written in restructured text and built with Sphinx, so it’s much easier to contribute
<Guest28>
I get that but the manual doesnt really talk about the python manifest like nothing what so ever I feel
<Guest28>
I dont care about the format so much, just that it shouldnt leave stuff out
<Guest28>
formats kind of a pain in the neck too, but I can live with that.
<moto_timo[m]>
This is a community right? Not a paid commercial product.
<Guest28>
I'll find some time and do that for sure. Just trying to figure all this out first. The whole manifest for python and the split packages. Not following it one single bit.
<Guest28>
Think I just need a break and get back to it in a bit. Sorry about the rant.
<moto-timo>
so if you know the path to the file... that tool helps
<moto-timo>
(also nees to be incorporated more into devtool/recipetool but time is finite)
<moto-timo>
"path to where the file would be installed on the target"
<moto-timo>
meh... I'm working on dunfell so python3.9 for anything new
<smurray>
heh, that bit me recently, some nice changes in 3.9 we can't use yet
<moto-timo>
I still think it is a valid doc issue... it just needs some attention and some time to do it write
<moto-timo>
And I do want to share the workflow that is almost automatic and lives in my head
LetoThe2nd has quit [Quit: Connection closed for inactivity]
<moto-timo>
(perl is even worse btw)
<Guest28>
can we back track to the oe-pkgdata-util find-path bit ..
<moto-timo>
s/write/right/
<moto-timo>
my English skills have attrophied horribly
<moto-timo>
oe-pkgdata-util is a script in meta/scripts
<Guest28>
im trying to run devtool to generate a recipe for python3-numbers , I just need like a place to get its source tree at I guess
<moto-timo>
it uses pkgdata that has been created by the build to know what packages provided what
<Guest28>
oe-okgdata-util would find the source tree from my local python distribution
<moto-timo>
ah... I might have thrown you off sorry
<Guest28>
like the python I got installed on my mac?
<moto-timo>
what I meant about python3-numbers reference is that the python3-manifest.json shows you that it is in a "special" (only for yp/oe) package called python3-numbers
<moto-timo>
so your mac or any normal distro would have the FULL python standard libraries installed in /usr/lib/python3.9/
<moto-timo>
or similar path
<moto-timo>
but we want the smallest possible image, so we split up the standard libraries into yp/oe specific sub-packages
<moto-timo>
those sub-packages are defined in python3-manifest.json
<moto-timo>
but oe-pkgdata-util is aware of those sub-packages (once it has built them)
<moto-timo>
"it" == bitbake
<Guest28>
oh ok, so itll get me like the path from the oe installation i got going on?
<moto-timo>
yup. it's another way to get the info
<Guest28>
and I can write a recipe using that as a SRC_URI or something
<moto-timo>
if... for instance... you arent' a huge fan of reading json
<moto-timo>
RDEPENDS_${PN} += "python3-numbers"
<moto-timo>
it's not source you are grabbing... it's a runtime dependency for the built package you are developing a recipe for
<moto-timo>
SRC_URI is for the tarball, or git repo, or the local files/patches for the specific recipe
<Guest28>
got it .. runtime dependency would just take of themselves as long as Ive got python on my image?
<Guest28>
or no! shoot not sure what I said back there
<Guest28>
I just need the runtime dependencies built out first to get the recipe which depends on them to build
<moto-timo>
if you install python3... I think rburton made that include the full standard library
<moto-timo>
because we confuse EVERYONE on the planet
<moto-timo>
but really the right way to do it is include ONLY what is needed
<moto-timo>
which is why recipetool/devtool start you with python3-core