Created attachment 220840 [details]
Increase maxmem of javadoc.
On amd64 this didn't fail, but arm64 fails.
http://ampere1.nyi.freebsd.org/data/121arm64-quarterly/557407/logs/errors/cassandra4-4.0.b1.log (ipv6 only)
http://ampere2.nyi.freebsd.org/data/head-arm64-default/p557699_s368500/logs/errors/cassandra4-4.0.b2.log (ipv6 only)
http://beefy18.nyi.freebsd.org/data/head-amd64-default/p558707_s368819/logs/cassandra4-4.0.b2.log (ipv6 only)
The increased memory on arm64 is probably because it does not use the hotspot JIT on openjdk8. Anyways, this small patch fixes it.
^Triage: remove patch Keyword.
We now derive whether or not a patch is contained via the [x] Patch checkbox in the diff itself. Using the metadata like this turns out to be a lot more consistent than relying on the manual effort to set the Keyword.
Comment on attachment 220840 [details]
testbuild @ work
(In reply to Kurt Jaeger from comment #3)
a testbuild fails with:
find: /bin/sh: Argument list too long
(In reply to Kurt Jaeger from comment #4)
It works for me.
So, where is the error?
My fix in this issue helped your and my build to continue past the javadoc stage so there is progress already.
The build error on your system is not on my RPI4 and not on any other arch which builds cassandra4 . I don't see how my patch would create this error.
In your log the "data seg size" 131072 in "--Resource limits--" seems a lot smaller than on my machine and on the amd64 beefy18 builder. Could that be the cause of "/bin/sh argument list too long"? But I'm guessing here. It was one of the small differences I notices when diffing your and my build log.
(In reply to Ronald Klop from comment #5)
I do not know where the resource limit is coming from or if raising it will fix the build. I'll ask on -arm@
(In reply to Kurt Jaeger from comment #6)
Hi Kurt, any news on the build failure in your qemu/arm64 setup?
I setup a RPI4 Jenkins server at home. It creates clean jails and runs poudriere in the clean jail.
Build fails without the patch and succeeds with the patch.
Anybody interested in committing this to fix the build on ARM64.
It has no regression on other platforms.
A commit references this bug:
Date: Fri Feb 5 19:53:56 UTC 2021
New revision: 564149
databases/cassandra4: fix aarch64/arm64 build (javadoc fails)
- Increase maxmem of javadoc
On amd64 the build does not fail, but on arm64 it fails.
Submitted by: Ronald Klop <email@example.com>
Approved by: firstname.lastname@example.org (maintainer)