unable to build mutter-46 for armv7 on x86_64 host

Cross compiling mutter 46 for armv7 on a x86_64 host fails with the following error below. Building mutter-45 for armv7 (e.g. mutter-mobile currently in pmaports) succeeds, but I suspect mutter-mobile will fail too once gnome-mobile is updated to work with mutter 46.

(process:56814): GLib-GObject-CRITICAL **: 18:33:05.900: validate_pspec_to_install: assertion 'G_IS_PARAM_SPEC (pspec)' failed
(process:56814): GLib-GObject-CRITICAL **: 18:33:05.901: g_param_spec_ref_sink: assertion 'G_IS_PARAM_SPEC (pspec)' failed
(process:56814): GLib-GObject-CRITICAL **: 18:33:05.901: g_param_spec_unref: assertion 'G_IS_PARAM_SPEC (pspec)' failed
qemu: uncaught target signal 11 (Segmentation fault) - core dumped
Command '['/home/pmos/build/src/mutter-46.1/output/tmp-introspectxw53akjt/Clutter-14', '--introspect-dump=/home/pmos/build/src/mutter-46.1/output/tmp-introspectxw53akjt/functions.txt,/home/pmos/build/src/mutter-46.1/output/tmp-introspectxw53akjt/dump.xml']' died with <Signals.SIGSEGV: 11>.

The actual failure is here:

[768/774] Generating clutter/clutter/Cally-14.gir with a custom command (wrapped by meson to set env)

I've been able to simplify reproducing it a bit by setting GI_SCANNER_DEBUG="save-temps" in mutter/APKBUILD, this lets us keep the tmp dir for g-ir-scanner (which seems to be failing AFAIK), and the failure can be quickly reproduced again after that by:

1. enter the chroot: pmbootstrap chroot -b armv7
2. su pmos, then cd /home/pmos/build/src/mutter-46.1/output (exact path from error message)
3. tmp-introspectxw53akjt/Clutter-14 --introspect-dump=tmp-introspectxw53akjt/functions.txt,tmp-introspectxw53akjt/dump.xml --> sigsegv

I'm unable to run it with gdb (something about ptrace not available?)...

Starting program: tmp-introspectxw53akjt/Clutter-14 --introspect-dump=tmp-introspectxw53akjt/functions.txt,tmp-introspectxw53akjt/dump.xml
warning: Could not trace the inferior process.
warning: ptrace: Function not implemented
During startup program exited with code 127.

When it fails under qemu-arm on x86_64, there's a message about qemu core dumping, but I haven't been able to figure out where this coredump is located: qemu: uncaught target signal 11 (Segmentation fault) - core dumped

Running it with valgrind shows that it goes off the rails when it tries to access memory at 0x4(?!). There are lots of "invalid write of size 4 errors" before this happens.

...
==2802== Invalid write of size 4
==2802==    at 0x49C10BE: ??? (in /usr/lib/libglib-2.0.so.0.8000.2)
==2802==  Address 0x3f8008a0 is on thread 1's stack
==2802==  16 bytes below stack pointer
==2802==

(process:2802): GLib-GObject-CRITICAL **: 01:15:55.221: g_param_spec_double: assertion 'default_value >= minimum && default_value <= maximum' failed

(process:2802): GLib-GObject-CRITICAL **: 01:15:55.507: validate_pspec_to_install: assertion 'G_IS_PARAM_SPEC (pspec)' failed

(process:2802): GLib-GObject-CRITICAL **: 01:15:55.529: g_param_spec_ref_sink: assertion 'G_IS_PARAM_SPEC (pspec)' failed
==2802== Invalid write of size 4
==2802==    at 0x4974AD2: g_param_spec_unref (in /usr/lib/libgobject-2.0.so.0.8000.2)
==2802==  Address 0x3f800b40 is on thread 1's stack
==2802==  8 bytes below stack pointer
==2802==

(process:2802): GLib-GObject-CRITICAL **: 01:15:55.541: g_param_spec_unref: assertion 'G_IS_PARAM_SPEC (pspec)' failed
==2802== Invalid read of size 4
==2802==    at 0x4970254: g_object_class_install_properties (in /usr/lib/libgobject-2.0.so.0.8000.2)
==2802==  Address 0x4 is not stack'd, malloc'd or (recently) free'd
==2802==
==2802==
==2802== Process terminating with default action of signal 11 (SIGSEGV)
==2802==  Access not within mapped region at address 0x4
==2802==    at 0x4970254: g_object_class_install_properties (in /usr/lib/libgobject-2.0.so.0.8000.2)

Mutter builds successfully for armv7 on x86_64 if crossdirect is disabled. For more info, see !5115 (comment 1899890119)