Closed nschonni closed 7 months ago
Yeah, here is the x86 logs and its definitely failing: https://unofficial-builds.nodejs.org/logs/202404241830-v22.0.0/x86.log
x86 looks broken and someone will need to look at it. Maybe the container and/or compiler will need to be updated, or there might be a real issue with x86 in V8.
I've requeued the remaining recipes:
./bin/queue-push.sh -v v20.0.0 -r musl -r x64-glibc-217 -r armv6l -r loong64 -r riscv64 -r x64-pointer-compression -r x64-debug
(FYI I'm not going to be around to babysit these through.)
Thanks @richardlau! I'll keep on eye on https://unofficial-builds.nodejs.org/logs/202404242126-v20.0.0/ and see if it works out
Oh, @richardlau I'm now seeing that you requeued v20 rather than v22. Not sure if that's going to cause an issue with the old images if the SHAs get updated for that old release
erk. Good catch. I've killed off the running processes/queued stuff and requeued for 22:
$ ./bin/queue-push.sh -v v22.0.0 -r musl -r x64-glibc-217 -r armv6l -r loong64 -r riscv64 -r x64-pointer-compression -r x64-debug
Queuing v22.0.0 with recipes: musl,x64-glibc-217,armv6l,loong64,riscv64,x64-pointer-compression,x64-debug
$
armv6l
uses a too old GCC
Although there may be architectures that are still failing, I'm going to close this one since the main issue of the build filing is addressed. If there are specific architecture issues that are going to persist in v22, it's probably good to spin up issues and disable them until they can be addressed
Isn't this going to be the case for the next v22 release though? should the x86 build be moved down the list to prevent this happening again?
https://github.com/nodejs/unofficial-builds/blob/main/recipes/armv6l/should-build.sh this should be fixed to stop it building >= 22, does someone here want to give that a go?
then, if someone wants to make this work again, they can put in the work to get the compiler upgraded, but as of today, it'd be on me and I'm very unlikely to do this any time soon
@rvagg yep! On it!
armv6l
uses a too old GCC
shouldn't armv6l
be disabled instead of x86
? @LaurentGoderre
@DanielSchuech the primary issue at this point is preventing the failing x86 from interfering with the next v22 release. I believe @rvagg was sending that file as an example on how to control the build.
Opened #143 for the armv6l by copying the other PR
Thanks for the explanation. I only had a look into the latest log folder where no x86.log
was present.
Thanks for the explanation. I only had a look into the latest log folder where no
x86.log
was present.
Yes, because I didn't queue x86 since it failed in the original attempt.
thanks for the two should-run changes @LaurentGoderre & @nschonni
If anyone wants to tackle armv6l, the path I'd take is to just produce a new https://github.com/rvagg/rpi-newer-crosstools with a the minimum compiler and libc needed to support node. It looks like there's newer crosstools-ng that support up to gcc 13 so it shouldn't be too hard. I think I explain the process in that repo about taking an existing config and running a newer crosstools-ng and picking new versions.
I think it ran the x86 build, then crashed out if I'm reading the logs right https://unofficial-builds.nodejs.org/logs/202404241830-v22.0.0/build.log