-
Notifications
You must be signed in to change notification settings - Fork 68
Segfault in sandmark frama-c #246
Comments
At ocaml-multicore/byterun/major_gc.c Lines 605 to 610 in 8fbacb1
there is a store to |
Frama-C assumes that it is compiled with naked-pointer support [0]. Multicore does not support naked pointers, nor does vanilla OCaml compiled with |
I think you can just allow zero as a distinguished null value, which is the current plan for upstream no-naked pointers, as I mentioned elsewhere. This is a small fix in the runtime, which I would suggest you make, so you can run these benchmarks. |
Do you mean to add |
I was planning for it to be a compile-time constant, so it can also be a compile-time constant in C stubs, and comparison against it is fast. |
While this sounds good, Frama-C really does need to be fixed. Frama-C is attempting to get an all-bits-zero null value by loading a word from |
Didn't realize that Frama-C was loading the wrong field! |
If you check out the
bugfix/ctk21/frama-c-multicore-effects
branch from @ctk21 on https://github.com/ocamllabs/sandmark and start a run with multicore the built frama-c segfaults.Backtrace:
The text was updated successfully, but these errors were encountered: