<Guest94>
Hi everyone. I'm facing a problem where I seems to struggle to fetch lfs files inside a submodule using gitsm. I'm on Scarthgap (using xilinx release 2024.2). When I check the file sizes of the lfs stored files in my compile step, they all are small (just the "pointer" files). I saw there was some recent work to fix this as it was broken in
<Guest94>
Landgdale, can it be that it's still broken?
Jones42 has joined #yocto
Guest94 is now known as psv
Jones42_ has joined #yocto
Jones42 has quit [Ping timeout: 252 seconds]
mbulut has quit [Remote host closed the connection]
psv has quit [Quit: Client closed]
Jones42 has joined #yocto
Jones42_ has quit [Ping timeout: 260 seconds]
jmiehe has joined #yocto
<derRichard>
i'm puzzled by the fitimage creation mechanism in yocto, it seems that there is very little control on how a fitimage it build. i'd like to define my own boot configs. e.g. one with and one without an initramfs
<derRichard>
do i miss something?
<rburton>
derRichard: there's a thread on the list about fit image generation, there more variables and hooks were being added. i was wondering if we just need to generalise it massively and instead of glueing it into kernel or uboot recipes, have a recipe class that just takes a .its source and dependencies and simply runs mkimage. your thoughts would be great.
* derRichard
looks
<rburton>
"uboot-sign - support ATF and TFE ITS generation"
<derRichard>
a recipe that builds a .its is exactly what i need
<rburton>
cool please reply :)
<rburton>
seemed like an easier solution than adding yet more variables into a complicated class
<rburton>
just a class that lets you have an .its in SRC_URI, depends on eg firmware and kernel, and it runs mkimage.
<derRichard>
is it on openembedded-core@?
<derRichard>
damn, i hate group.io so much
<rburton>
yes and yes
<derRichard>
new member moderated, so i can't post nor see messages
<Emantor>
derRichard: there is also fitimage.bbclass in meta-oe.
davidinux has quit [Quit: WeeChat 4.3.1]
aardo has quit [Ping timeout: 252 seconds]
<derRichard>
Emantor: sure? never saw that one. i also don't see such a file. i'm on scarthgap
<Emantor>
derRichard: unfortunately it only arrived in meta-oe on styhead, but a backport is trivial.
<derRichard>
Emantor: let me check :)
<derRichard>
rburton: thx, i found the thread with google. but still need to find a way how to reply
<ak77>
hmm... i am rusty around edges... ERROR: Nothing PROVIDES 'networkmanager-nmtui-native'. Close matches: networkmanager-native RPROVIDES networkmanager-nmtui-native ? is having PACKAGES += "..." not enough ?
<rburton>
Emantor: ah that's interesting. do you find that better than "just write an its file"?
<derRichard>
Emantor: this class sounds cool! i'll give it a try
<Emantor>
rburton: We do, generating its files can be pretty complicated if you need to support i.e. multiple boards.
<rburton>
i didn't notice that
<rburton>
you should have submitted it to core ;)
jmiehe has quit [Quit: jmiehe]
<Emantor>
rburton: You'll need to argue with enrico about that :)
<ak77>
i need nmtui on host, and will be connecting to networkmanager via dbus on a target.
<ak77>
I already do that, but there is a version mismatch, and would like to build proper nmtui native
<rburton>
ak77: the "proper" way considering -native recipes are pretty limited by design would be build a nativesdk version and use that
<derRichard>
Emantor: how is this class supposed to be used? i thought i can inhert it to my rootfs image. but i get circular dependency on do_populate_mfgtool
<rburton>
derRichard: separate recipe?
<ak77>
rburton: ok. what about... just getting nmtui out of networkmanager recipe
florian has joined #yocto
<ak77>
networkmanager-nmtui-native that is
<rburton>
you might find some of the dependencies don't exist in native and there'll be resistance to add them
Jones42 has quit [Ping timeout: 244 seconds]
<derRichard>
rburton: ah. sure :D
<derRichard>
stupid me
davidinux has joined #yocto
Tyaku has quit [Quit: leaving]
davidinux has quit [Quit: WeeChat 4.3.1]
davidinux has joined #yocto
psv has joined #yocto
<psv>
Sorry, I lost connection for a minute there. Unfortunately I don't have any history here in the web client, so if someone was so kind to answer me, I didn't see it. :-(
ehussain has quit [Remote host closed the connection]
rfuentess has quit [Read error: Connection reset by peer]
rfuentess has joined #yocto
Jones42 has joined #yocto
davidinux has quit [Quit: WeeChat 4.3.1]
davidinux has joined #yocto
<Emantor>
derRichard: separate recipe is the way.
<sakoman>
rburton: it required a bit of tweaking for kirkstone, but I now have it in my test queue for kirkstone and scarthgap
<rburton>
sakoman: might be worth testing if you have a 2404 machine to hand, not sure if theres anything else needed with that
psv has quit [Quit: Client closed]
<sakoman>
rburton: I don't. Will autobuilder testing on 2404 suffice?
<rburton>
sakoman: i can test tomorrow if you mail me to remind me :)
<sakoman>
rburton: OK, will do!
davidinux has quit [Quit: WeeChat 4.3.1]
ardo has joined #yocto
davidinux has joined #yocto
<derRichard>
Emantor: is there a reason why i have to define loadaddress and entrypoint manually?
<Emantor>
derRichard: Hm, at least for barebox it's not a requirement.
<Emantor>
Doesn't U-Boot choose a suitable entrypoint and loadaddress if none is given?
* Emantor
has to confess that he very rarely has to touch U-Boot.
<derRichard>
Emantor: i have to confess that i'm using an old u-boot
<derRichard>
maybe recent u-boot does something sane
<derRichard>
at least here, u-boot hates the generated fit images beacause load/entry addrs are missing
davidinux has quit [Quit: WeeChat 4.3.1]
Xagen has quit [Ping timeout: 244 seconds]
davidinux has joined #yocto
cyxae has joined #yocto
cyxae has quit [Remote host closed the connection]
cyxae has joined #yocto
Daanct12 has quit [Quit: WeeChat 4.4.4]
Jones42_ has joined #yocto
Jones42 has quit [Ping timeout: 244 seconds]
rob_w has quit [Quit: Leaving]
Xagen has joined #yocto
steelswords94 has quit [Quit: Ping timeout (120 seconds)]
steelswords94 has joined #yocto
jmiehe has joined #yocto
halloy4743 has quit [Ping timeout: 260 seconds]
Noor has joined #yocto
halloy4743 has joined #yocto
<Noor>
Hello guys, I need a recommendation from you guys. we have our own distro layer. So, what does community say regarding changing PACKAGECONFIG. Should it be changed in bbappend for a particular recipe or it should be done in ditro file? Both will work but what is the recommendation
goliath has quit [Quit: SIGSEGV]
jmiehe has quit [Quit: jmiehe]
psv has joined #yocto
psv has quit [Client Quit]
psv has joined #yocto
<psv>
Trying my luch again :)
<psv>
Hi everyone. I'm facing a problem where I seems to struggle to fetch lfs files inside a submodule using gitsm. I'm on Scarthgap (using xilinx release 2024.2). When I check the file sizes of the lfs stored files in my compile step, they all are small (just the "pointer" files). the problem with submodules fetching in general, but I'm still not able to make it fetch LFS inside a submodule.
<psv>
I saw there was some recent work to fix this as it was broken in Langdale, and has since gotten a fix. This seems to solve
<psv>
the submodule fetching, but not LFS fetching in my case.
ClaytonC has joined #yocto
frgo has quit [Remote host closed the connection]
frieder has quit [Remote host closed the connection]
rcw has joined #yocto
Noor has quit [Quit: Client closed]
ben-m-z has quit [Ping timeout: 244 seconds]
<psv>
The issue seems to be an surplus "--no-fetch" in unpack line 255.
florian has quit [Remote host closed the connection]
<psv>
in gitsm.py
frgo has joined #yocto
rfuentess has quit [Remote host closed the connection]
prabhakalad has quit [Ping timeout: 276 seconds]
prabhakalad has joined #yocto
ClaytonC has quit [Quit: Client closed]
frgo has quit [Remote host closed the connection]
rstreif has joined #yocto
<rber|res>
@derRicher, @Emantor: as far as I remember if you use a uImage the entrypoints need to be defined UBOOT_ENTRYPOINT, UBOOT_LOADADDRESS, you can define them as 0, then you need to pass them from the u-boot environment or so
<rber|res>
@derRichard ^^^
frgo has joined #yocto
florian has joined #yocto
zpfvo has quit [Quit: Leaving.]
<rber|res>
Let's say you build a core-image-minimal and a core-image-sato-sdk from sources (no sstate) on 2 different machines and you would like to __merge__, __combine__ the 2 sstates. Is this possible? How?
<rber|res>
Just copy one over the other?
jackkush has joined #yocto
<jackkush>
yo
jackkush has left #yocto [#yocto]
frgo has quit [Ping timeout: 246 seconds]
<derRichard>
rber|res: sure sure. so did it. i assumed to get this automatically like the kernel-fitimage bbclass does
<rber|res>
@derRichard: I am also not sure how you can automatically determine where load the kernel and the device tree, it depends on the kernel size and the device tree size so they don't overlap
<rber|res>
@derRichard: it's something like ramstart + 0x8000 for the kernel if I remember well
<rber|res>
@derRichard: so it's also hardware dependent
florian has quit [Ping timeout: 244 seconds]
jmd has joined #yocto
pbiel has quit [Remote host closed the connection]
pbiel has joined #yocto
ben-m-z has joined #yocto
ben-m-z has quit [Quit: ben-m-z]
davidinux has quit [Read error: Connection reset by peer]
olku has joined #yocto
<olku>
hi there. Is it possible to disable build of kernel and uboot ( temporary )?
<derRichard>
rber|res: well, kernel-fitimage.bbclass gets it from the bsp layer. e.g. meta-freescale sets UBOOT_ENTRYPOINT, etc.
<vmeson>
olku: yes, there's a way to just build userspac for a container image for example. Let me check for the variable...
<vmeson>
olku: seems ot be: PREFERRED_PROVIDER_virtual/kernel = "linux-dummy"
frgo has joined #yocto
leon-anavi has quit [Quit: Leaving]
prabhakalad has quit [Ping timeout: 246 seconds]
prabhakalad has joined #yocto
frgo has quit [Ping timeout: 265 seconds]
<rber|res>
@derRichard: yes the same for all others
florian has joined #yocto
olku has quit [Quit: Client closed]
iceaway has quit [Quit: Client closed]
iceaway has joined #yocto
iceaway has quit [Client Quit]
<AdrianF>
derRichard: we set UBOOT_ENTRYPOINT and UBOOT_LOADADDRESS in the machine.conf file. And we use KERNEL_CLASSES += "kernel-fitimage".
<derRichard>
AdrianF: yep, that's a common approach. all i'm saying is that fitimage.bbclass (not to be confused with kernel-fitimage.bbclass) should also use these variables
olku has joined #yocto
Kubu_work has quit [Quit: Leaving.]
druppy has joined #yocto
<KanjiMonster>
also the dtb address in ram has a 8-byte alignment requirement, which isn't guaranteed within the fitimage, so you need to set a dtb loadaddr either in u-boot or the fitimage itself
frgo has joined #yocto
<AdrianF>
derRichard: fitimage.bbclass is something your are working on?
<AdrianF>
derRichard: Interesting. This approach can potentially simplify things a lot in comparison to using the kernel-fitimage.bbclass. I was not aware that this already exists in meta-oe. My learning was that the kernel recipes should not support fitImage at all. It should provide the kernel and an independent recipe should then pack it into a fitImage.
<AdrianF>
Was on my todo list to try this. But now I see this. Thanks.
pbiel has quit [Ping timeout: 252 seconds]
<derRichard>
AdrianF: i learned about that class today
<khem>
derRichard:maybe rsysnc would work good for merging the sstates
<derRichard>
khem: ?
pivi has joined #yocto
ptsneves has joined #yocto
<khem>
derRichard: I was looking through old msgs and did you ask about merging sstate cache from two different builds
pivi has left #yocto [#yocto]
<derRichard>
khem: ah. that must have been a long time ago :D
<derRichard>
and yes, i'm using rsync these days :)
goliath has joined #yocto
frgo has joined #yocto
<khem>
psv: do you have git-lfs installed on build machine ?
<khem>
derRichard: it was in my scrollback now I cant even scroll that far back so I guess it could be old
frgo has quit [Ping timeout: 244 seconds]
Xagen has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
olku has quit [Quit: Client closed]
ptsneves has quit [Ping timeout: 252 seconds]
cyxae has quit [Quit: cyxae]
<rber|res>
@khem: this was my question with sstates
<khem>
rber|res:oh ok :)
florian has quit [Ping timeout: 260 seconds]
<khem>
tab completion foxed me then
<rber|res>
I am torturing myself with CI and gitea act_runner, which refuses to run longer than 3 hours - I think they have a bug somewhere. 3 places where you can change the timeout and it stays 3 hours ;) So I split my jobs in chunks less then 3 hours - each job produces it's own SSTATE
<zeddii>
khem. looks like vboxguestdrivers (as usual) blows up with the new kernel. I'll take a look at it tomorrow.
<zeddii>
but zfs and turbostat (two others that blow up) are building. so might just be that for the most part.