Deactivate memory tracking once freed

[krtab]

No description provided.

[zapashcanon]

I'm not sure I understand this one. Could you explain when such a pointer is going to be used?

Also, I'm surprised the code in src/libc/stdlib.c is not changed. Shouldn't you change realloc and free along the way ?

[filipeom]

Looks good to me!

Just cast the calls to owi_free to void to get rid of the import type errors.

For example, change free in

owi/src/libc/src/stdlib.c

Line 50 in aef0bd6

void free(void *ptr) { owi_free(ptr); }

To

void free(void *ptr) { (void)owi_free(ptr); }

[filipeom]

I'm not sure I understand this one. Could you explain when such a pointer is going to be used?

The rust allocator does a bunch of offset reads with a heap chunk when deallocating stuff. So in order to avoid having false positive heap overflows e when deallocating things in rust we need dealloc to deregister the pointer from Owi's chunk table and return an untracked pointer so that the rust allocator can do nasty stuff with it.

Also, I'm surprised the code in src/libc/stdlib.c is not changed. Shouldn't you change realloc and free along the way ?

free and realloc can remain the same I think

[zapashcanon]

I see. It would be great if we could start to test these things in Owi itself...

@krtab there is a conflict that needs to be resolved before I can merge

[filipeom]

I see. It would be great if we could start to test these things in Owi itself...

What do you suggest? I think for this case in particular it may be a bit difficult since we don't have the rust frontend atm

[zapashcanon]

I'm not suggesting anything for the current PR, but yes, it would be nice to have a small Rust frontend integrated in Owi for us to test this.

[zapashcanon]

Thanks!