wasmtime / issue #7037 Memory checker (wmemcheck) complet... · git-wasmtime

Stream: git-wasmtime

Topic: wasmtime / issue #7037 Memory checker (wmemcheck) complet...

Wasmtime GitHub notifications bot (Sep 14 2023 at 17:44):

This issue is meant to track what remains to make wmemcheck feature-complete and robust enough for practical use:

[ ] Handle SIMD loads and stores (#6829)

[ ] Hook calloc as well as existing malloc and free

[ ] Provide hooks for custom allocators to use

[ ] Allow for more flexibility in heap and stack layout (don't assume stack is above .data/.bss and everything below initial TOS is accessible)

[ ] Provide an alternate way to hook malloc/free, rather than requiring a name section in the .wasm module

[ ] Provide an option to print warnings as violations occur, rather than halting execution (like Valgrind)

[ ] Provide a mechanism to define "suppressions" for unwanted warnings (like Valgrind)

[ ] Track definedness at byte or bit level?

cc @iximeow for more thoughts, and others from today's Wasmtime meeting?

Wasmtime GitHub notifications bot (Sep 15 2023 at 15:39):

iximeow commented on issue #7037:

@alexcrichton noted earlier that it would be good to have:

[ ] an end-to-end test starting with some clearly-UB .c and test expectations for a stack trace pointing to wasmtime trapping a bad access

and since wmemcheck doesn't support multiple memories (though we ought to make it so!),

[ ] warn if a user tries to wmemcheck a multi-memory program
or add that support multiple memories. but i think that's maybe a step after "robust enough for practical use" in terms of effort, warning seems more expedient for now :)

Wasmtime GitHub notifications bot (Sep 15 2023 at 15:40):

iximeow edited a comment on issue #7037:

@alexcrichton noted earlier that it would be good to have:

[ ] an end-to-end test starting with some clearly-UB .c and test expectations for a stack trace pointing to wasmtime trapping a bad access

and since wmemcheck doesn't support multiple memories (though we ought to make it so!),

[ ] warn if a user tries to wmemcheck a multi-memory program

or add that support multiple memories. but i think that's maybe a step after "robust enough for practical use" in terms of effort, warning seems more expedient for now :)

Last updated: Nov 22 2024 at 16:03 UTC