Fix issue #34101 #34109
Fix issue #34101 #34109
Conversation
|
r? @jroesch (rust_highfive has picked a reviewer for you, use r? to override) |
|
r? @arielb1 |
|
This patch was meant to be a somewhat minimal change to the code in question, at the potential cost of wasting compilation time evaluating I am assuming that the recursive chain there almost never gets terribly long in most (human-generated) code. But its worth keeping an eye on, and I'm open to alternative suggestions. |
|
I would rather have that logic inside Maybe also add a soft assertion that all non- r+ modulo that |
| /// | ||
| /// Lvalues behind ADT's with a Drop impl are not tracked by | ||
| /// elaboration since they can never have a drop-flag state that | ||
| /// differs fom that of the parent with the Drop impl. |
|
@arielb1 oh good: I think moving the logic into |
…nor gather flags for untracked content. (Includes a regression test, which needed to go into `compile-fail/` due to weaknesses when combining `#[deny(warnings)]` with `tcx.sess.span_warn(..)`) (updated with review feedback from arielb1.)
|
@bors r=arielb1 |
|
📌 Commit 4b6a68e has been approved by |
|
hmm actually I think this patch might break |
|
In particular: run-pass/issue-4401.rs regresses (when compiled without (I also saw a failure associated with |
Fix issue rust-lang#34101 Fix issue rust-lang#34101: do not track subcontent of type with dtor nor gather flags for untracked content. (Includes a regression test, which needed to go into `compile-fail/` due to weaknesses when combining `#[deny(warnings)]` with `tcx.sess.span_warn(..)`)
Fix issue #34101: do not track subcontent of type with dtor nor gather flags for untracked content.
(Includes a regression test, which needed to go into
compile-fail/due to weaknesses when combining
#[deny(warnings)]withtcx.sess.span_warn(..))