09:37 <JaMa> https://lists.openembedded.org/g/bitbake-devel/message/17414 will save one nuclear power plant worth of electricity in builds with meta-qti :)

16:03 <vmeson> JaMa: meta-qti-bsp or meta-qt[456] or some other layer that I can't find?

16:16 <smurray> vmeson: former, I believe

16:22 <JaMa> yes, meta-qti (not the public one)

16:23 <JaMa> it uses all weird and not-recommended ways to break your builds (including AUTOREV for https://git.codelinaro.org/clo which just had planned maintenance https://status.codelinaro.org/ so all the builds failed

16:26 <JaMa> and instead of having sources fetched by bitbake fetcher based on SRC_URI it unpack own prebuilt files and ends with checksums of many files in do_fetch tasks and from 10 builds I did yesterday on different servers I got 9 different checksums because the ordering was different (which was fixed in bitbake in nanbield, but never backported to kirkstone - because most people don't use so many checksums and with

16:27 <JaMa> externalsrc you probably don't really expect the same src dir to be identical on all the builders

16:29 <JaMa> and unfortunate side effect of this bug is that the sstate signature of these do_fetch tasks is different (and filename is different), but bitbake-diffsigs doesn't show why, only that the task hash is different

16:56 <vmeson> JaMa: smurray: thanks. Sounds like "fun!"

17:02 <JaMa> indeed, but I've seen some good people starting to work for them, so it can only get better (I hope)

17:03 <JaMa> they do good hw, but bsp is even worse than amlogic