Seen on Ubuntu 20.04 with gcc 10.5.0 when building jdk24 (now mainline), and I have a report of this on Fedora 40 with GCC 14 building jdk 23.

Build dies after lto processes get OOM-killed. I see two `lto1` processes, each with an RSS between 20 and 30 GB. The rise in rss is very sharp.

```
PID RSS S TTY TIME COMMAND
485860 27077096 R pts/3 00:01:12 /usr/lib/gcc/x86_64-linux-gnu/10/lto1 -quiet -dumpdir /shared/projects/openjdk/jdk-jdk/output-release/support/modules_libs/java.base/server/ -dumpbase libjvm.so.wpa -m64 -mtune=generic -march=x86-64 -auxbase abs tractCompiler -g -O3 -O3 -fno-openmp -fno-openacc -fPIC -fcf-protection=full -fuse-linker-plugin -fno-strict-aliasing -fltrans-output-list=/tmp/libjvm.so.jBP7cv.ltrans.out -fwpa=24 -fresolution=/tmp/ccUfqx3v.res -flinker-output=dyn @/tmp/cc0FswT y
485897 23165692 R pts/3 00:01:10 /usr/lib/gcc/x86_64-linux-gnu/10/lto1 -quiet -dumpdir /shared/projects/openjdk/jdk-jdk/output-release/hotspot/variant-server/libjvm/gtest/ -dumpbase libjvm.so.wpa -m64 -mtune=generic -march=x86-64 -auxbase gtest Main -g -O3 -O3 -fno-openmp -fno-openacc -fPIC -fcf-protection=full -fuse-linker-plugin -fno-strict-aliasing -fltrans-output-list=/tmp/libjvm.so.tt4n2i.ltrans.out -fwpa=24 -fresolution=/tmp/ccV4dz4U.res -flinker-output=dyn @/tmp/ccMgrI5k
```

Relevant configure options:

--with-gtest=/shared/projects/openjdk/gtest/googletest-1.14.0 --with-debug-level=release --disable-precompiled-headers --with-jvm-features=link-time-opt

Not sure if we can do anything about this. I mainly open the bug to track this problem.

Note that I also see asserts firing during building in fastdebug, with the JVM complaining that HandleMark are not on the current thread stack. Maybe lto optimizes away stack-bound RAII objects? But note that lto1 crashes here, not the JVM.