Don't free embedded, unsized slices in their dtor #16838
Conversation
|
LGTM, modulo s/vectors/slices/. EDIT: oops, I didn't realize you were actually handling |
| @@ -63,16 +63,11 @@ pub fn make_drop_glue_unboxed<'a>( | |||
|
|
|||
| let len = get_len(bcx, vptr); | |||
| let dataptr = get_dataptr(bcx, vptr); | |||
| let bcx = if ty::type_needs_drop(tcx, unit_ty) { | |||
| if ty::type_needs_drop(tcx, unit_ty) { | |||
There was a problem hiding this comment.
It looks like prior to your merge, the if statement actually covered the len and dataptr vars too. I don't know if that really matters.
if ty::type_needs_drop(tcx, unit_ty) {
let fill = get_fill(bcx, vptr);
let dataptr = get_dataptr(bcx, vptr);
iter_vec_raw(bcx, dataptr, unit_ty, fill, glue::drop_ty)
} else { bcx }There was a problem hiding this comment.
Actually, this is calling get_fill() instead of get_len(). Curious.
There was a problem hiding this comment.
iter_vec_raw was changed to take the length of the vec, not the fill (len*size_of_element) and get_fill was replaced with get_len
There was a problem hiding this comment.
the len and dataptr could/should be inside the if
Thanks to @eddyb for finding the bug. Closes rust-lang#16826 (I hope)
|
Changed approach, r? |
pnkfelix
changed the title
|
This failure does not look related, but is worrisome in general. I have very occasionally seen this happen locally (run-pass hanging). I'm going to do a cc @alexcrichton -- have you seen these sorts of hangs before? |
|
Yes bors has be failing quite a bit "spuriously" recently, I sadly haven't had time to investigate just yet :( |
Don't double free embedded, unsized slices.
Merge/rebase error from DST. Thanks to @eddyb for finding.
Closes #16826 (I hope)
r?