On Mon, Nov 29, 2021 at 07:36:30AM -0500, Kaleb Keithley wrote:
> This seems to be happening on a semi-regular basis now. I run scratch
> builds and they invariably work, but I have seen at least one of my scratch
> builds fail with the same error on f36/rawhide.
<snip>
/builddir/build/BUILD/ceph-16.2.6/src/os/bluestore/BlueStore.cc: In member function 'void BlueStore::_kv_sync_thread()':
/builddir/build/BUILD/ceph-16.2.6/src/os/bluestore/BlueStore.cc:12280:1: internal compiler error: maximum number of generated reload insns per insn achieved (90)
12280 | }
| ^
Yes, that's the error I've seen too.
I'd suggest filing a bug against GCC IN Fedora. The GCC maintainers are
typically quick to respond and help diagnose such issues.
I note that where it started failing saw a systemtap rebase. This is a
bit of a flag for me, because last week we saw glusterfs builds start
failing with internal compiler errors due this systemtap rebase. That
was in arm7, while your failures are x86, but still makes me wonder
if your build problems go away if you test with disabled systemtap
Ceph doesn't seem to have an option to disable systemtap.
I will file a BZ against gcc.
Thanks and regards,