Re: [PATCH] c++: further lookup_member simplification

[Jason Merrill via Gcc-patches]

On 3/15/22 13:18, Patrick Palka wrote:

As a followup to r12-7656-gffe9c0a0d3564a, this minor patch condenses
the handling of ambiguity and access w.r.t. the value of 'protect' so
that it more clearly matches the function comment.

Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
trunk?

OK.

gcc/cp/ChangeLog:

        * search.cc (lookup_member): Simplify by handling all values
        of protect at once in case of ambiguous lookup.  Don't modify
        protect.
---
  gcc/cp/search.cc | 32 +++++++++++++-------------------
  1 file changed, 13 insertions(+), 19 deletions(-)

diff --git a/gcc/cp/search.cc b/gcc/cp/search.cc
index 85e3e7cb487..b86b3a24080 100644
--- a/gcc/cp/search.cc
+++ b/gcc/cp/search.cc
@@ -1168,27 +1168,21 @@ lookup_member (tree xbasetype, tree name, int protect, 
bool want_type,
    if (rval_binfo)
      type = BINFO_TYPE (rval_binfo);

- /* If we are not interested in ambiguities, don't report them;

-     just return NULL_TREE.  */
-  if (!protect && lfi.ambiguous)
-    return NULL_TREE;
-
-  if (protect == 2)
-    {
-      if (lfi.ambiguous)
-       return lfi.ambiguous;
-      else
-       protect = 0;
-    }
-
-  if (protect == 1 && lfi.ambiguous)
+  if (lfi.ambiguous)
      {
-      if (complain & tf_error)
+      if (protect == 0)
+       return NULL_TREE;
+      else if (protect == 1)
        {
-         error ("request for member %qD is ambiguous", name);
-         print_candidates (lfi.ambiguous);
+         if (complain & tf_error)
+           {
+             error ("request for member %qD is ambiguous", name);
+             print_candidates (lfi.ambiguous);
+           }
+         return error_mark_node;
        }
-      return error_mark_node;
+      else if (protect == 2)
+       return lfi.ambiguous;
      }

if (!rval)

@@ -1213,7 +1207,7 @@ lookup_member (tree xbasetype, tree name, int protect, 
bool want_type,

only the first call to "f" is valid. However, if the function is

      static, we can check.  */
-  if (protect && !really_overloaded_fn (rval))
+  if (protect == 1 && !really_overloaded_fn (rval))
      {
        tree decl = is_overloaded_fn (rval) ? get_first_fn (rval) : rval;
        decl = strip_using_decl (decl);