<ptsneves>
I was trying to get the poky test passing by removing the cleanall but i have a dirty workspace and other things got in the way. When i have time today i will try to finish fixing the test.
<RP>
ptsneves: there are a few fixes in master-next but the big worry is that testsdkext is failing on eSDKs
<RP>
ptsneves: I ran out of time last night so I'm trying to reproduce this morning
<ptsneves>
do you have the builderbot logs or was it locally?
<RP>
ptsneves: I still don't have that test that got broken working right either
<ptsneves>
@RP sent the fix to the openembedded's test_invalid_recipe_src_uri. It is one more line than what you have in the master-next. I tested it is sucessfull
<RP>
ptsneves: would you be interested/able to tweak it a little bit more? I'd like to get the "The paths that were searched were:" in this message
florian_kc has joined #yocto
<ptsneves>
@RP sure. I need to add the signoff either way
<RP>
ptsneves: that is a local thing, we can't add that for you
<LetoThe2nd>
ptsneves: that would implicitly assume that everybody who sends a patch is authorized to, instead of requiring people to explicitly acknowledge it.
Guest5 has quit [Quit: Client closed]
kami has quit [Quit: Client closed]
<ptsneves>
LetoThe2nd: indeed, but if a user sends something by sharing he is already doing a deliberate action of sharing. Regardless it is not technically feasible. I had the wrong impression that client side hooks were trackeable by git
<LetoThe2nd>
ptsneves: the difference is between "sharing information" and "being authorized to convey this information to some other project"
<rburton>
I can email out a confidential document but that doesn't mean I was authorised to :)
Habbie has quit [Ping timeout: 244 seconds]
<RP>
ptsneves: it seems we use DL_DIR in there for file:// to file:// url mappings in MIRRORS
<ptsneves>
@RP cant we add DL_DIR to the FILESPATH where those mappings are expected, localizing that logic to the state cache?
<ptsneves>
even if the end result does not end up being very different from the current one.
<RP>
ptsneves: it is in FILESPATH :/
<RP>
ptsneves: I don't understand where the issue is as yet, just that it is the file:// to file:// mirror code
Guest5 has joined #yocto
<ptsneves>
Unable to get checksum for man-db SRC_URI entry 404: file could not be found
<ptsneves>
RP: Will try to have a look later. In the mean time i corrected the oeqa patch to point to the correct mailinglist(oe-devel) instead of bitbake-devel and added the signedoff. For the bitbake patches i added the paths searched to the errors and also fixed a missing signoff. The bbfatal looks something like:
<RP>
ptsneves: great, that makes it as useful as the current fetch failure now
<RP>
ptsneves: thanks, I'll continue to try and see if I can work out what is going wrong with sstate
ardo has quit [Read error: Connection reset by peer]
ardo has joined #yocto
Habbie has joined #yocto
wkawka has quit [Quit: Client closed]
nemik has quit [Ping timeout: 240 seconds]
nemik has joined #yocto
nemik has quit [Ping timeout: 276 seconds]
nemik has joined #yocto
wkawka has joined #yocto
<wkawka>
Hi, how can I trace wic blupgin? I have created a new plugin from the old one, but my created grub.cnf file is being overwritten, and I can't find anything what is doing that
<wkawka>
plugin*
<RP>
ptsneves: I think this is actually a missing mkdir :/
<ptsneves>
@RP can you point to the code you are looking at?
<ptsneves>
localpath used to point to the last entry of localpaths which was DL_DIR
<ptsneves>
or better it is DL_DIR in current code.
<RP>
ptsneves: I think it was just luck it usually worked. This might not fix all the issues, not sure yet. Will need another run. It certainly makes the logs quieter
<ptsneves>
I think the DL_DIR fallback was providing a very pervasive fallback even in cases where it was not meant to.
rob_w has quit [Remote host closed the connection]
<LetoThe2nd>
I have a Makefile-based package that with INIT_MANAGER = "systemd" borks out on ld not understanding -Wl,-O1 whereas on "sysvinit" it builds, and I totally can't find why. Can't find a difference in the LDFLAGS and TARGET_LDFLAGS. any pointers?
kscherer has joined #yocto
<rburton>
LetoThe2nd: have you compared the log.do_compiles with and without systemd?
pabigot has quit [Remote host closed the connection]
<RP>
or even the configure logs
sakoman has joined #yocto
alicef has quit [Quit: install gentoo]
alicef has joined #yocto
alicef has quit [Client Quit]
<LetoThe2nd>
heh, tried to do that now (thanks rburton), so did a cleansstate and tried to reproduce. guess what happens? recipe builds for systemd.
alicef has joined #yocto
pabigot has joined #yocto
<rburton>
bare makefile recipes are notoriously bad. it might be rebuilds which break it
<rburton>
if you can, implement a working 'make clean'
alicef has quit [Quit: install gentoo]
alicef has joined #yocto
<qschulz>
or just delete the recipe tmpdir between tries also?
<rburton>
that's cheating if the goal is to solve the problem
<LetoThe2nd>
who cares about solving....
<qschulz>
rburton: it was the way to get the log.do_compile/do_configure back?
<qschulz>
I assume the WORKDIR was dirty and make didn't re-run completely
<qschulz>
and if LetoThe2nd is using the same workdir for his tries with INIT_MANAGER, it might mess up the attempts?
<qschulz>
much knowledgeable, very expert
* LetoThe2nd
much depressed, very drink
<wkawka>
During booting my system hangs on systemd-journald[1843]: Received client request to flush runtime journal.
<wkawka>
It is hanging for abiut 90 seconds, then i can run shell for maintenance
<rburton>
LetoThe2nd: if this is something you wrote, just use a proper build tool. if its something else, shout at the author to use a proper build tool
<LetoThe2nd>
rburton: i will save my breath instead of shouting at some chinese author writing boot0 for allwinner, tbh.
<qschulz>
rburton: I think we just closed the loop :) what you said is the first thing LetoThe2nd always say when someone asks something about makefile based projets :D
<agherzan>
1. it backports a patch so ninja will adhere to the cgroup limits
<agherzan>
2. it implements a way for injecting a load average configuration via an environment variable (this can be set in the CI infra and/or we can discuss a sane default for oe-core)
<agherzan>
We use 2 for recipes doing bare ninja stuff, for gn-based recipes but ideally it would be used in cmake too (to pass the relevant argument down to ninja). Do these look sane for pushing to master/kirkstone.
<agherzan>
?
elroy has joined #yocto
<elroy>
Hi. I find a lot of text in the Yocto manual about running GDB on the target. Are there any tips for using GDB on the host to analyze postmortem core dumps sent back by users?
<JaMa>
agherzan: looks nice (I would be definitely interested to test this), but doesn't "export MAXLOAD_NINJA" screw with sstate reuse?
<ptsneves>
JaMa: Would this not be good for the patches on bitbake for the cpu pressure?
<agherzan>
Yes, it would and it will. I'll need to fix that JaMa
<JaMa>
agherzan: if you happen to add it in bitbake.conf's BB_HASHEXCLUDE_COMMON then please remove PARALLEL_MAKE from BB_HASHCONFIG_IGNORE_VARS as it's listed in both variables (and the later includes BB_HASHEXCLUDE_COMMON so it's listed twice)
<vmeson>
RP: agherzan: -- reading -- cgroups eh? no root perms needed even to set things up?
<JaMa>
RP: yes, you have to use -O for scp to work
<JaMa>
RP: I've seen your message about -O being only temporary, but I haven't found any confirmation from upstream about it
<landgraf>
JaMa: with 'PACKAGE_EXCLUDE_COMPLEMENTARY openssh-sftp-server will be installed so you can use scp still.
<agherzan>
vmeson: that is only used to use the existing cgroup limits. It doesn't set them
<agherzan>
That is only to use*
<agherzan>
It basically sets the -j based on the cgroup limit.
<vmeson>
agherzan: cool. That would certainly help. Instead of load average, we've found that /proc/pressure/cpu regulation using either avg10 or a manual 1 second diff of the total pressure is more responsive.
<landgraf>
so I've found this less invasive than bad_recommendation. the only disadvantage is missed openssh-ptest/-dev/-dbg package on openssh-based images with *-pkgs but this can be simply worked around with IMAGE_INSTALL:append = "openssh-<pkg>"
<vmeson>
agherzan: I'm in another meeting so I'll skim now and read more carefully later.
<agherzan>
vmeson: but this will throttle ninja directly as it spawns new jobs
<agherzan>
How are you using the CPU pressure?
adams[1] has quit [Ping timeout: 252 seconds]
<agherzan>
Oh I see, you are only saying that this will be a better metric
<vmeson>
agherzan: so far only to limit bitbake from starting new tasks but I think the same approach would be useful for ninja/make. Let me find the link....
<vmeson>
agherzan: yes.
<agherzan>
We would need to implement it in make/ninja etc
<agherzan>
The limit on bitbake will be just the tip of the iceberg because we need the throttling for the invoked builds too
<vmeson>
agherzan: agree. I can share some of the work we've done so far if you like. Setting PARALLEL_MAKE to a small value of say 1/Number-of-builds does help limit the overload of a full buld but won't help with chromium for example.
<agherzan>
I would start for now with the existing support in make and ninja and we can improve it with PSI afterwards
<vmeson>
along with the /proc/pressure regulation that is.
<agherzan>
Chromium was cmake?
<agherzan>
If yes, I plan to extend the ave load conf to cmake too
<vmeson>
agherzan: sure, it'll help. I did test using ninja -l with chromium and as you'd expect the number of compilation jobs shot above and below the targetted limit but it did help on average.
manuel1985 has quit [Ping timeout: 244 seconds]
* vmeson
shifts focus back to the meeting he's in...
<agherzan>
We'll run some tests on our farm. And tell you how it goes
<fenrig>
is there a way to do it only for aarch64?
<JaMa>
PACKAGECONFIG:remove:aarch64
<fenrig>
can I do it from wherever?
<fenrig>
this structure?
<qschulz>
fenrig: in a bbappend for the recipe
<fenrig>
I only have it when combining glibc2.35 with <linux5.4 on aarch64
<fenrig>
so I would like to try to disable the feature only there :/
florian_kc has quit [Ping timeout: 240 seconds]
AkkermT_ has joined #yocto
<qschulz>
fenrig: I don't think there's a guaranteed way to know which version of the kernel will be built from within recipes
<fenrig>
too bad, I put a comment around it. As Im mainly focussing on linux containers which will be running on 4.14 or greater :/
nerdboy has quit [Ping timeout: 244 seconds]
<fenrig>
qschulz: btw isnt the linux-libc-headers component an implicit dependency of all the packages that are to be built for the target? as its part of the toolchain
<JaMa>
we work around this by specifying the kernel version to build in MACHINE.conf and then it can be used in other recipes, but that's not without its own issues as well
<fenrig>
btw I dont have any kernel built, but I do need the headers in the cross compile toolchain to match the host
<fenrig>
to clarify, the host of the container
jpuhlman_ has joined #yocto
jpuhlman has quit [Killed (zirconium.libera.chat (Nickname regained by services))]
<janvermaete[m]>
What to do? Email to mailinglist. A patch? But how to verify if the recipe is still working. I'm not using the recipe itself.
florian_kc has joined #yocto
amitk has quit [Ping timeout: 244 seconds]
<zeddii>
RP: I have -stable updates, and then the next set of buildpaths fixes. They'd all go on top of my last ones (gen-mach-types). Want me to do a pull request, versus sending the 6 patches as singles .. that way the ordering is clear
<Xagen>
i have an odd issue that i'm hoping someone can help solve
<Xagen>
i have a binary that's failing to run due to it saying that it doesn't exist (but does)
<Xagen>
and it looks like one of the libraries is the cause
<Xagen>
when i use ldd on it from devshell, i see "/lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.34' not found"
<rfried>
Hi. I'm trying to add customized kernel (git) to a new BSP I'm creating.
<rfried>
The build fails with: ERROR: Nothing RPROVIDES '${@oe.utils.conditional('KERNEL_IMAGETYPE',' (but /work/ramon/bsp/bsp/yocto/layers/meta-neureality/recipes-kernel/linux/linux-nr_5.18.bb RDEPENDS on or otherwise requires it)
<Xagen>
but glibc 2.35 is what should be getting installed based on the recipes i see
<kanavin_>
back in berlin! but not starting to work until 25th july or so
<Xagen>
i think it's looking at my host glibc for some reason
<Xagen>
not sure how to get around this
<rfried>
Would love if someone can point me to what am I missing. because KERNEL_IMAGETYPE is defined in the machine config.
adams[1] has joined #yocto
<JPEW>
rsalveti: After talking with RP, we (well, mostly RP) found a much simpler solution to the SPDX document problem we were seeing
manuel1985 has quit [Ping timeout: 272 seconds]
<JPEW>
I just pushed the patch to the ML
nemik has quit [Ping timeout: 268 seconds]
nemik has joined #yocto
<JPEW>
RP: it could also be `if isSPDXtask(task) and isSPDXtask(deptaskname)` so that it only applies to do_create_(runtime_)spdx -> do_create_(runtime_)spdx`. I can't decide if that's better
nemik has quit [Ping timeout: 272 seconds]
nemik has joined #yocto
<jonmason>
probably a stupid question, but any idea why pulseaudio would need m4-native with a binary toolchain but not native gcc?
manuel1985 has joined #yocto
<jonmason>
doing an armv7 build of sato and seeing this (building sato for the first time for this platform, so unsure if this is a regression or not)
manuel1985 has quit [Ping timeout: 260 seconds]
manuel1985 has joined #yocto
<rsalveti>
JPEW: awesome, let me take a look
manuel1985 has quit [Ping timeout: 240 seconds]
<rsalveti>
ah, cool, changing sstatesig
manuel1985 has joined #yocto
<dmoseley>
At the meeting in Nuremburg for EW, there was some discussion of using Matrix with IRC integration for accessing these channels. Does anyone have any links or docs on how to do this? Also, what would be the benefit of that setup over an IRC bouncer like ZNC? cc: LetoThe2nd