Auto merge of #118082 - compiler-errors:rollup-ejsc8yd, r=matthiaskrgr

Rollup of 8 pull requests

Successful merges:

 - #117828 (Avoid iterating over hashmaps in astconv)
 - #117832 (interpret: simplify handling of shifts by no longer trying to handle signed and unsigned shift amounts in the same branch)
 - #117891 (Recover `dyn` and `impl` after `for<...>`)
 - #117957 (if available use a Child's pidfd for kill/wait)
 - #117988 (Handle attempts to have multiple `cfg`d tail expressions)
 - #117994 (Ignore but do not assume region obligations from unifying headers in negative coherence)
 - #118000 (Make regionck care about placeholders in outlives components)
 - #118068 (subtree update cg_gcc 2023/11/17)

r? `@ghost`
`@rustbot` modify labels: rollup

1 files changed, 7 insertions, 2 deletions

diff --git a/compiler/rustc_codegen_ssa/src/base.rs b/compiler/rustc_codegen_ssa/src/base.rs
index 198e5696357..23054975548 100644
--- a/compiler/rustc_codegen_ssa/src/base.rs
+++ b/compiler/rustc_codegen_ssa/src/base.rs
@@ -322,8 +322,13 @@ pub fn cast_shift_expr_rhs<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
     if lhs_sz < rhs_sz {
         bx.trunc(rhs, lhs_llty)
     } else if lhs_sz > rhs_sz {
-        // FIXME (#1877: If in the future shifting by negative
-        // values is no longer undefined then this is wrong.
+        // We zero-extend even if the RHS is signed. So e.g. `(x: i32) << -1i8` will zero-extend the
+        // RHS to `255i32`. But then we mask the shift amount to be within the size of the LHS
+        // anyway so the result is `31` as it should be. All the extra bits introduced by zext
+        // are masked off so their value does not matter.
+        // FIXME: if we ever support 512bit integers, this will be wrong! For such large integers,
+        // the extra bits introduced by zext are *not* all masked away any more.
+        assert!(lhs_sz <= 256);
         bx.zext(rhs, lhs_llty)
     } else {
         rhs