[Bug middle-end/103364] s390x: TLS reference in /usr/lib64/libLLVM.so mismatches non-TLS reference in /usr/lib64/libLLVM.so

["sarah.kriesch at opensuse dot org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103364

--- Comment #18 from Sarah Julia Kriesch <sarah.kriesch at opensuse dot org> ---
(In reply to Andreas Krebbel from comment #17)
> (In reply to Sarah Julia Kriesch from comment #12)
> > that is happening during the build process in OBS with a really minimal
> > openSUSE Tumbleweed. We are using VMs using QEMU and with 4GB of memory.
> 
> Why only 4GB? Isn't this way too low for building things like rust with lto
> and everything?
> 
We are sharing the LinuxONE together with SUSE. Not only OBS is using it. It is
also used for automated tests with openQA. Therefore, the resources are kept
minimal for build processes. The highest amount of memory for openSUSE based
builds is 12GB. I had to argue at the beginning of my time for s390x, that we
can receive it. Rust is using 8GB split into two 4GB processes. That is given
by a constraints file. The OBS is choosing a free worker node matching the
requirements.
But also 8GB did not help for PostgreSQL14. We are receiving the same error
message. And about the topic "too slow": x86 is using the same at PostgreSQL14
and is much faster than s390x for most builds.

> I've successfully built rust1.54 and postgresql14 several times in an
> opensuse tumbleweed container. So I would suspect either the kernel or the
> guest setup you are using. Could it perhaps be that ld processes got oom
> killed and have left half-complete binaries which triggered the error then?
> 
I checked the container images. They are updated continuously. Therefore, that
should be reproducible with that (if you are using my list with packages incl.
LLVM13). The kernel version can not be the issue. We are receiving this error
since multiple kernel versions and only with s390x.
I have checked the infrastructure with a build in the LinuxONE Community Cloud.
I am receiving the same issue with following commands:

sudo zypper in osc build obs-service-format_spec_file bsdtar #also possible
with other Linux distributions
osc co openSUSE:Factory:zSystems/postgresql14
cd openSUSE\:Factory\:zSystems/postgresql14/
osc build --vm-type=kvm --vm-memory=4G

> In the current logs I don't see the ld issue anymore. Apparently you already gave it more memory and the behavior changed due to that?
That was another issue between. The issue came back this weekend.
I tested it already with 8GB. I will test it also with 12GB in my mainframe VM
over night now. But I do not expect any change there.