From mboxrd@z Thu Jan  1 00:00:00 1970
From: Claude Code Review Bot <claude-review@example.com>
To: dri-devel-reviews@example.com
Subject: Claude review: gpu: nova-core: emit Rust metadata for dependent modules
Date: Thu, 04 Jun 2026 16:23:29 +1000
Message-ID: <review-patch3-20260530-nova-exports-v3-3-1202aa339ef7@nvidia.com>
In-Reply-To: <20260530-nova-exports-v3-3-1202aa339ef7@nvidia.com>
References: <20260530-nova-exports-v3-0-1202aa339ef7@nvidia.com> <20260530-nova-exports-v3-3-1202aa339ef7@nvidia.com>
X-Mailer: Claude Code Patch Reviewer
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

Patch Review

**Status: Looks good**

This adds Makefile rules to produce `libnova_core.rmeta` so dependent Rust =
modules can use `--extern nova_core=3D<path>` for compile-time type resolut=
ion.

The build is isolated in `.nova_core_rmeta/` to avoid racing with the main =
`nova_core.o` compilation on rustc's intermediate file names =E2=80=94 this=
 is a real problem since rustc writes intermediate objects with determinist=
ic names. The implementation:

```makefile
cmd_rustc_rmeta_rs =3D mkdir -p $(dir $@); \
    $(rust_common_cmd) --emit=3Dmetadata=3D$@ \
    --emit=3Dobj=3D$(basename $@).o $<; \
    rm -f $(basename $@).o
```

The `--emit=3Dobj=3D...` followed by `rm -f` is necessary because some rust=
c versions require code generation to produce correct metadata. The discard=
ed `.o` is placed in the private directory so it doesn't collide.

The two-step approach (private build =E2=86=92 `if_changed,copy` to stable =
path) ensures consumers see a stable path and the copy only happens when co=
ntent changes. The `private __modname`, `part-of-builtin`, and `part-of-mod=
ule` settings correctly replicate the build context.

The downside is compiling nova-core effectively twice (once for .o, once fo=
r .rmeta), but this is acceptable for a temporary workaround.

---
Generated by Claude Code Patch Reviewer