public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc(refs/vendors/ARM/heads/morello)] cp: Various __intcap and __null fixes
@ 2022-03-14 10:35 Matthew Malcomson
  0 siblings, 0 replies; only message in thread
From: Matthew Malcomson @ 2022-03-14 10:35 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:8eadbc50a134aa1e6dd465d91cdc9878d930ff33

commit 8eadbc50a134aa1e6dd465d91cdc9878d930ff33
Author: Alex Coplan <alex.coplan@arm.com>
Date:   Tue Feb 8 15:56:13 2022 +0000

    cp: Various __intcap and __null fixes
    
    GCC gives __null an INTCAP_TYPE when Pmode is a capability mode: this
    follows CHERI LLVM.
    
    The initial problem here was c_common_type_for_mode building new intcap
    type nodes instead of using the shared (u)intcap_type_node nodes. This
    meant these type nodes didn't go through the usual builtin type
    registration and consequently didn't have a TYPE_NAME. This in turn
    meant we called c_pretty_printer::simple_type_specifier on these types
    and hit ICEs due to the TYPE_PRECISION checks on capability types (as
    well as emitting sub-par diagnostics involving __null). This patch fixes
    both of these issues by:
     - Using the shared tree nodes in c_common_type_for_mode.
     - Adjusting the precision checks to use TYPE_CAP_PRECISION.
    
    Note we use c_common_type_for_mode (ptr_mode, 0) in
    c-common.c:c_common_nodes_and_builtins to build null_node (used for
    __null in C++, NULL is defined to this). Hence the need to fix
    c_common_type_for_mode when called with CADImode here.
    
    In this patch, we also implement various conversions that were found to
    be needed with __null having INTCAP_TYPE (and of course are needed for
    __intcap in general). We also adjust null_ptr_cst_p to recognize null pointer
    constants of INTCAP_TYPE.
    
    We adjust tsubst to handle substituting into an INTCAP_TYPE node (i.e.
    (u)intcap_type_node) by simply returning the node unmodified. It looks
    like the INTEGER_TYPE case is specifically for handling substituting
    into the TYPE_DOMAIN of an ARRAY_TYPE, so we needn't worry about that
    handling applying to INTCAP_TYPEs (the TYPE_DOMAIN of an array type is
    guaranteed to be an INTEGER_TYPE, and we intend to preserve that
    guarantee).
    
    gcc/c-family/ChangeLog:
    
            * c-common.c (c_common_type_for_mode): Use shared
            (u)intcap_type_node nodes instead of building a new one.
            * c-pretty-print.c (c_pretty_printer::simple_type_specifier):
            Adjust a couple of TYPE_PRECISION uses to TYPE_CAP_PRECISION.
    
    gcc/cp/ChangeLog:
    
            * call.c (null_ptr_cst_p): Recognize null pointer constants of
            INTCAP_TYPE: this is the type of __null under pure/fake-cap.
            (standard_conversion): Allow implicit conversion from INTCAP_TYPE to
            arithmetic types.
            * cvt.c (cp_convert_to_pointer): Implement __intcap -> pointer
            conversion.
            * pt.c (tsubst): Handle INTCAP_TYPE.
            * typeck.c (build_reinterpret_cast_1): Allow reinterpret_casting
            __intcap to pointer.

Diff:
---
 gcc/c-family/c-common.c       | 2 +-
 gcc/c-family/c-pretty-print.c | 4 ++--
 gcc/cp/call.c                 | 5 +++--
 gcc/cp/cvt.c                  | 3 +++
 gcc/cp/pt.c                   | 1 +
 gcc/cp/typeck.c               | 3 ++-
 6 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/gcc/c-family/c-common.c b/gcc/c-family/c-common.c
index 650b1c94610..dd0ded25d9d 100644
--- a/gcc/c-family/c-common.c
+++ b/gcc/c-family/c-common.c
@@ -2266,7 +2266,7 @@ c_common_type_for_mode (machine_mode mode, int unsignedp)
       else
 	{
 	  gcc_assert (is_a <scalar_addr_mode> (mode));
-	  return build_intcap_type_for_mode (mode, unsignedp);
+	  return unsignedp ? uintcap_type_node : intcap_type_node;
 	}
     }
 
diff --git a/gcc/c-family/c-pretty-print.c b/gcc/c-family/c-pretty-print.c
index 87e96845e2e..ddc29c67464 100644
--- a/gcc/c-family/c-pretty-print.c
+++ b/gcc/c-family/c-pretty-print.c
@@ -350,7 +350,7 @@ c_pretty_printer::simple_type_specifier (tree t)
 	}
       else
 	{
-	  int prec = TYPE_PRECISION (t);
+	  int prec = TYPE_CAP_PRECISION (t);
 	  tree common_t;
 	  if (ALL_FIXED_POINT_MODE_P (TYPE_MODE (t)))
 	    common_t = c_common_type_for_mode (TYPE_MODE (t),
@@ -361,7 +361,7 @@ c_pretty_printer::simple_type_specifier (tree t)
 	  if (common_t && TYPE_NAME (common_t))
 	    {
 	      simple_type_specifier (common_t);
-	      if (TYPE_PRECISION (common_t) != prec)
+	      if (TYPE_CAP_PRECISION (common_t) != prec)
 		{
 		  pp_colon (this);
 		  pp_decimal_int (this, prec);
diff --git a/gcc/cp/call.c b/gcc/cp/call.c
index c0c52120e14..d083808b877 100644
--- a/gcc/cp/call.c
+++ b/gcc/cp/call.c
@@ -545,12 +545,12 @@ null_ptr_cst_p (tree t)
       /* Core issue 903 says only literal 0 is a null pointer constant.  */
       if (TREE_CODE (t) == INTEGER_CST
 	  && !TREE_OVERFLOW (t)
-	  && TREE_CODE (type) == INTEGER_TYPE
+	  && (TREE_CODE (type) == INTEGER_TYPE || INTCAP_TYPE_P (type))
 	  && integer_zerop (t)
 	  && !char_type_p (type))
 	return true;
     }
-  else if (CP_INTEGRAL_TYPE_P (type))
+  else if (CP_INTEGRAL_TYPE_P (type) || INTCAP_TYPE_P (type))
     {
       t = fold_non_dependent_expr (t, tf_none);
       STRIP_NOPS (t);
@@ -1485,6 +1485,7 @@ standard_conversion (tree to, tree from, tree expr, bool c_cast_p,
   else if (ARITHMETIC_TYPE_P (to) || INTCAP_TYPE_P (to))
     {
       if (! (INTEGRAL_CODE_P (fcode)
+	     || fcode == INTCAP_TYPE
 	     || (fcode == REAL_TYPE && !(flags & LOOKUP_NO_NON_INTEGRAL)))
           || SCOPED_ENUM_P (from))
 	return NULL;
diff --git a/gcc/cp/cvt.c b/gcc/cp/cvt.c
index a36a3cce8af..182646674d3 100644
--- a/gcc/cp/cvt.c
+++ b/gcc/cp/cvt.c
@@ -243,6 +243,9 @@ cp_convert_to_pointer (tree type, tree expr, bool dofold,
       return error_mark_node;
     }
 
+  if (form == INTCAP_TYPE)
+    return convert_to_pointer_maybe_fold (type, expr, dofold);
+
   if (INTEGRAL_CODE_P (form))
     {
       if (capability_type_p (type))
diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
index 5dbdd37f6e3..afea77c84c9 100644
--- a/gcc/cp/pt.c
+++ b/gcc/cp/pt.c
@@ -15333,6 +15333,7 @@ tsubst (tree t, tree args, tsubst_flags_t complain, tree in_decl)
     case BOOLEAN_TYPE:
     case NULLPTR_TYPE:
     case LANG_TYPE:
+    case INTCAP_TYPE:
       return t;
 
     case INTEGER_TYPE:
diff --git a/gcc/cp/typeck.c b/gcc/cp/typeck.c
index 571e5ab36a9..c467cdbbfdf 100644
--- a/gcc/cp/typeck.c
+++ b/gcc/cp/typeck.c
@@ -7995,7 +7995,8 @@ build_reinterpret_cast_1 (location_t loc, tree type, tree expr,
   /* [expr.reinterpret.cast]
      A value of integral or enumeration type can be explicitly
      converted to a pointer.  */
-  else if (TYPE_PTR_P (type) && INTEGRAL_OR_ENUMERATION_TYPE_P (intype))
+  else if (TYPE_PTR_P (type) && (INTEGRAL_OR_ENUMERATION_TYPE_P (intype)
+				 || INTCAP_TYPE_P (intype)))
     /* OK */
     ;
   else if ((INTEGRAL_OR_ENUMERATION_TYPE_P (type)


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-03-14 10:35 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-03-14 10:35 [gcc(refs/vendors/ARM/heads/morello)] cp: Various __intcap and __null fixes Matthew Malcomson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).