Fix suggestion span involving wrongly placed generic arg on enum variants

When the variant and the (wrongly placed) args are at vastly separated
source locations such as being in different macos or one in a macro and
the other somwhere outside of it, the arg spans we computed would
span the entire distance between such locations and were hence invalid.
.

Author: gurry

compiler/rustc_hir_analysis/src/astconv/mod.rs

@@ -36,7 +36,7 @@ use rustc_middle::ty::{
 use rustc_session::lint::builtin::AMBIGUOUS_ASSOCIATED_ITEMS;
 use rustc_span::edit_distance::find_best_match_for_name;
 use rustc_span::symbol::{kw, Ident, Symbol};
-use rustc_span::{sym, Span, DUMMY_SP};
+use rustc_span::{sym, BytePos, Span, DUMMY_SP};
 use rustc_target::spec::abi;
 use rustc_trait_selection::traits::wf::object_region_bounds;
 use rustc_trait_selection::traits::{self, NormalizeExt, ObligationCtxt};
@@ -1275,8 +1275,10 @@ impl<'o, 'tcx> dyn AstConv<'tcx> + 'o {
                                 return;
                             };
                             // Get the span of the generics args *including* the leading `::`.
-                            let args_span =
-                                assoc_segment.ident.span.shrink_to_hi().to(args.span_ext);
+                            // We do so by stretching args.span_ext to the left by 2. Earlier
+                            // it was done based on the end of assoc segment but that sometimes
+                            // led to impossible spans and caused issues like #116473
+                            let args_span = args.span_ext.with_lo(args.span_ext.lo() - BytePos(2));
                             if tcx.generics_of(adt_def.did()).count() == 0 {
                                 // FIXME(estebank): we could also verify that the arguments being
                                 // work for the `enum`, instead of just looking if it takes *any*.