Hi,

h8300-hms-g++ fails to compile some C++ files, for example when building
brickos (0.9.0.dfsg-15) demo:

/usr/bin/h8300-hms-g++ -DCXX -fno-rtti -fno-exceptions -O2 -fno-builtin \
	-fomit-frame-pointer -Wall \
	-I/build/reproducible-path/package/include \
	-I/build/reproducible-path/package/include/lnp \
	-I. \
	-I/build/reproducible-path/package/boot \
	-c lightSensor.C -o lightSensor.o
In file included from lightSensor.C:19:
/build/reproducible-path/package/include/c++/LightSensor.H:46: internal compiler error: Segmentation fault

This seems to happen since gcc-h8300-hms is built with gcc-15.

This bug may be the same that the one that has been seen previously by
Matthew Sheets, when trying to run the GCC tests after being build with
GCC-15.

Nicolas.

Issues stemming from GCC 15 seem to be surfacing elsewhere, too.

Consider the packages overview for the Debian LEGO team.
 * https://qa.debian.org/developer.php?email=debian-lego-team%40lists.alioth.debian.org

Package gcc-h8300-hms shows Buildd failing on sh4.

Taking a closer look at that, and it was okay until comparatively recently:
 * https://buildd.debian.org/status/logs.php?pkg=gcc-h8300-hms&arch=sh4

Log from the last "successful" build on 2025-04-10:
 * https://buildd.debian.org/status/fetch.php?pkg=gcc-h8300-hms&arch=sh4&ver=1%3A3.4.6%2Bdfsg2-7&stamp=1744322988&raw=0

Log from the next subsequent failed build on 2025-09-07:
 * https://buildd.debian.org/status/fetch.php?pkg=gcc-h8300-hms&arch=sh4&ver=1%3A3.4.6%2Bdfsg2-8&stamp=1757254675&raw=0


In each of those logs, scroll down to the "Build environment" section and note
a key difference in the "Toolchain package versions" listed for each:
 * Succeeding 2024-04-10 build:  GCC 14
 * Failing 2025-09-07 build:  GCC 15

Furthermore, the error given for the 2025-09-07 build failure is an
Internal Compiler Error:
```
qemu: uncaught target signal 11 (Segmentation fault) - core dumped
gcc: internal compiler error: Segmentation fault signal terminated program cc1
Please submit a full bug report, with preprocessed source (by using -freport-bug).
See <file:///usr/share/doc/gcc-15/README.Bugs> for instructions.
```


For the issues I've observed (referenced by the preceding message in this bug
report), I have listed some of my findings at the link below.
 * https://github.com/BrickBot/GNU-Legacy-Toolchain/issues/14

Briefly from that GitHub link, here is a short code snippet that will fail compilation:
``` C
struct S { ~S(); };
void bar();
void foo()
{
  S s;
  bar();
}
```


If I drop the optimization level from -O2 to -O1, the problems encountered
when using GCC 15 to build GCC 3.4.6 seem to go away.


Matthew

Failures upon cutover from GCC 14 to GCC 15 also occurred with the related
package binutils-h8300-hms on Hurd (both i386 and amd64):
 * https://qa.debian.org/developer.php?email=debian-lego-team%40lists.alioth.debian.org

Package binutils-h8300-hms shows Buildd failing on hurd-amd64 and hurd-i386:
 * https://buildd.debian.org/status/package.php?p=binutils-h8300-hms

Again, both were generally okay until comparatively recently:
 * hurd-amd64
   - https://buildd.debian.org/status/logs.php?pkg=binutils-h8300-hms&arch=hurd-amd64
 * hurd-i386
   - https://buildd.debian.org/status/logs.php?pkg=binutils-h8300-hms&arch=hurd-i386

Logs from the last "successful" build (2025-04-02)
 * hurd-amd64
   - https://buildd.debian.org/status/fetch.php?pkg=binutils-h8300-hms&arch=hurd-amd64&ver=2.16.1-13&stamp=1743617394&raw=0
 * hurd-i386
   - https://buildd.debian.org/status/fetch.php?pkg=binutils-h8300-hms&arch=hurd-i386&ver=2.16.1-13&stamp=1743630118&raw=0

Logs from the next subsequent failed build (2025-09-06)
 * hurd-amd64
   - https://buildd.debian.org/status/fetch.php?pkg=binutils-h8300-hms&arch=hurd-amd64&ver=2.16.1-14&stamp=1757201855&raw=0
 * hurd-i386
   - https://buildd.debian.org/status/fetch.php?pkg=binutils-h8300-hms&arch=hurd-i386&ver=2.16.1-14&stamp=1757199085&raw=0

In both cases (amd64 and i386), GCC 15 is _not_ part of the build environment for the 2025-04-02 builds
but _is_ part of the build environment for the 2025-09-06 builds.  (The hurd-i386 environment on
2025-09-06 is actually reporting that _both_ GCC 15 and GCC 14 are installed.)


Matthew