diff options
| author | Paul Buetow <paul@buetow.org> | 2026-05-29 22:38:23 +0300 |
|---|---|---|
| committer | Paul Buetow <paul@buetow.org> | 2026-05-29 22:38:23 +0300 |
| commit | 8da6c62f617d6351a4ab6062bf2ea104b3b31ec2 (patch) | |
| tree | 417797280abcbaa7946b1ad28043d63a895a2b81 /internal/generate/codegen_test.go | |
| parent | 18c8e8f5f3d7cc9bbbcb9b8b9be65477110363a7 (diff) | |
test(generate): lock in get_robust_list/set_robust_list classification
Audit of get_robust_list(2)/set_robust_list(2) found the existing
classification already correct and consistent across classify.go,
family.go, the generated C handlers, and docs/syscall-tracing-plan.md:
- enter is KindNull: args[0] of get_robust_list is a PID, not an fd,
and head_ptr/len_ptr are userspace output pointers (no fd/path), so
the pid must not be picked up as a file descriptor.
- exit is ret_event UNCLASSIFIED: both return 0/-1 with no byte count.
- family is Misc, grouped with the per-thread sibling rseq rather than
promoted to IPC like the futex_* shared-memory primitives.
Add TestClassifyGetRobustListPidNotFd pinning these invariants (Kind,
Family, Ret) plus a futex_* IPC contrast case, mirroring the prior
sched_getattr/recvmsg audit lock-in tests. No behavior change.
The IPC-vs-Misc family question for the robust-list pair is tracked as
a separate follow-up task, not changed here (no clear correctness case).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Diffstat (limited to 'internal/generate/codegen_test.go')
0 files changed, 0 insertions, 0 deletions
