Refactor note on preferred poison-test approach

Signed-off-by: David Horstmann <david.horstmann@arm.com>
This commit is contained in:
David Horstmann 2023-10-23 19:53:30 +01:00
parent 52df620736
commit 806055edbf

View File

@ -501,10 +501,12 @@ This should either poison or unpoison the given buffer based on the value of `sh
* When `should_poison == 1`, this is equivalent to calling `VALGRIND_MAKE_MEM_NOACCESS(buffer, length)` or `ASAN_POISON_MEMORY_REGION(buffer, length)`.
* When `should_poison == 0`, this is equivalent to calling `VALGRIND_MAKE_MEM_DEFINED(buffer, length)` or `ASAN_UNPOISON_MEMORY_REGION(buffer, length)`.
We may choose one of two approaches. As discussed in [the design exploration](#validation-with-existing-tests), the first is preferred:
We may choose one of two approaches:
* Use transparent allocation-based memory poisoning.
* Use memory poisoning functions and a new testsuite.
As discussed in [the design exploration](#validation-with-existing-tests), the transparent approach is preferred.
We will specify the particularities of each approach's implementation below.
#### Transparent allocation-based memory poisoning