vasu-the-sharma wrote:

> I am also a bit confused about what your test is checking. Is it verifying 
> the lack of support from the `null` and `aligned` sanitizers? I see very few 
> `CHECK` lines looking for sanitizer intrinsics, just a bunch of memcpy ops 
> which are emitted even when the sanitizers are off.
> 
> Is your test documenting a gap or testing a functionality of these sanitizers?

This test is meant to document the current behavior as a baseline, so that when 
we implement `UBSAN` null/alignment checks for aggregate copies in a follow-up 
PR, we can update the test accordingly and verify the before/after changes.

The idea is:
- **This PR (NFC):** Establish a baseline document that memcpy is emitted but 
no sanitizer checks exist yet
- **Follow-up PR:** Implement the actual `UBSAN` instrumentation for aggregate 
copies, then update this test to check for 
`__ubsan_handle_type_mismatch_v1_abort` calls before the `memcpy`

https://github.com/llvm/llvm-project/pull/176210
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to