01:13 <paulg> 1: rust-1.80.1-r0 do_install - 11m20s (pid 13823)

01:13 <paulg> 2: cargo-c-0.10.5+cargo-0.83.0-r0 do_compile - 11m10s (pid 14188) 73% |##################################################################################################### |

01:13 <paulg> 3: llvm-19.1.6-r0 do_compile - 9m57s (pid 20290) 9% |##############

01:14 <paulg> and what do you know - even a 32G machine OOMs and barfs

01:14 <paulg> (bitbake world)

01:15 <paulg> We shouldn't ever be attempting all three of those at the same damn time.

01:15 <paulg> Friggin monsters.

04:04 <khem> paulg: yeah and sadly rust is compiling is an internal version of llvm there :)

04:05 <paulg> ugh.

04:05 <khem> I think bootstrapping distro with build from source is becoming harder and harder for desktop form favtor

04:05 <khem> factor

04:06 <khem> others do not see this issue as much since they use prebuilt versions either binary packages or intermediate cache ( nixos )

04:07 <khem> but problem is real. someone recently told me that rebuilding fedora rawhide can go for days ( RISCV h/w in this case )

04:08 <khem> but the pain is taken by a small set of build machines and pain is not fanned to every user/developer

04:08 <khem> they can use the prebuilt once done

04:09 <khem> I have seen memory around 43G in use when llvm/rust/cargo compile in parallel

04:10 <khem> I wish we had a way to share sstate like torrents

04:12 <khem> I was hoping bitbake's pressure meter would have toned down the parallelism and contained the load

11:27 <ak77> so, image build fails while doing do_rootfs, it fails on opkg, "Unable to update the package index files", it is unable to create temp dir inside '${T}/ipktemp' (only mentioned in scarthgap/meta/lib/oe/package_manager/ipk/__init__.py#L181), as there is no ipktemp dir. who should create this ipktemp directory ?

11:34 <ak77> (using kas, scrathgap, DL_DIR and SSTATE_DIR redirected).

11:37 <ak77> '${T}/ipktemp' is used as -t temp dir for opkg. but when running through kas (same layers) it's not created, when using bitbake directly (same _DIR redirections) it it somehow created. don't know by who though

11:43 <pbiel> I'm trying to use https://git.yoctoproject.org/poky/plain/meta-selftest/recipes-test/images/test-empty-image.bb to make an empty image, however it does not seem to work. My image still contains /etc/ bin /lib/ /sbin /usr and /var directories. Is there a way to completly get rid of them?

11:52 <Guest77> Hello together, Does anyone have any idea how to get debug information in binaries in Yocto Scrapthgap? It used to be present in my yocto project before upstreaming to Scrapthgap under /tmp/work..../packages-split/. I shall appreciate your response

11:59 <rburton> guest77: work/ is transient so won't be there is a recipe was built from sstate. install the -dbg packages on the target (or just use the dbg-pkgs image feature to install them all) to get the debug symbols on the target, or create a debugfs for remote gdb usage, or use debuginfod. see the documentation for all of those.

12:31 <rburton> pbiel: did you try the image i pointed you at last week?

12:32 <pbiel> Yep but it didn't help. However I managed to fix it just a few minutes before. I removed all the directories adding a command to IMAGE_PREPROCESS_COMMAND

12:33 <pbiel> It's rather workaround not the ideal solution but is enough for now

13:04 <ak77> any ideas?

13:09 <ak77> it's strange. man(7) for mkdtemp() says it returns ENOENT if directory in given pathname is missing, but only place where "ipktemp" is mentioned is in that package_manager's ipk __init__ file. so who creates that directory. I have checked opkg source and it does not create all paths (like mkdir -p).