C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Marek Polacek]

This ought to fix the fallout from PR c/68162 fix.  Here the problem is that
grokdeclarator created a wrong type for PARM_DECL "p".  It created this decl
with type "const int[<unknown>] *" while it should be "const int *".

I think the problem is that we weren't using TREE_TYPE on orig_qual_type and
thus c_build_qualified_type and subsequent c_build_pointer_type might create
a bogus type.  So when we're transfering const-ness of an array into that of
type pointed to, use TREE_TYPE not only of "type", but even of the orig qual
type.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-12-03  Marek Polacek  <polacek@redhat.com>

	PR c/68668
	* c-decl.c (grokdeclarator): When creating a PARM_DECL of ARRAY_TYPE,
	use TREE_TYPE of orig_qual_type.

	* gcc.dg/pr68668.c: New test.

diff --git gcc/c/c-decl.c gcc/c/c-decl.c
index 9ad8219..0edff2a 100644
--- gcc/c/c-decl.c
+++ gcc/c/c-decl.c
@@ -6417,6 +6417,8 @@ grokdeclarator (const struct c_declarator *declarator,
 	  {
 	    /* Transfer const-ness of array into that of type pointed to.  */
 	    type = TREE_TYPE (type);
+	    if (orig_qual_type != NULL_TREE)
+	      orig_qual_type = TREE_TYPE (orig_qual_type);
 	    if (type_quals)
 	      type = c_build_qualified_type (type, type_quals, orig_qual_type,
 					     orig_qual_indirect);
diff --git gcc/testsuite/gcc.dg/pr68668.c gcc/testsuite/gcc.dg/pr68668.c
index e69de29..d144fb6 100644
--- gcc/testsuite/gcc.dg/pr68668.c
+++ gcc/testsuite/gcc.dg/pr68668.c
@@ -0,0 +1,10 @@
+/* PR c/68668 */
+/* { dg-do compile } */
+
+typedef const int T[];
+
+int
+fn1 (T p)
+{
+  return p[0];
+}

	Marek

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Joseph Myers]

On Thu, 3 Dec 2015, Marek Polacek wrote:

> This ought to fix the fallout from PR c/68162 fix.  Here the problem is that
> grokdeclarator created a wrong type for PARM_DECL "p".  It created this decl
> with type "const int[<unknown>] *" while it should be "const int *".
> 
> I think the problem is that we weren't using TREE_TYPE on orig_qual_type and
> thus c_build_qualified_type and subsequent c_build_pointer_type might create
> a bogus type.  So when we're transfering const-ness of an array into that of
> type pointed to, use TREE_TYPE not only of "type", but even of the orig qual
> type.

I think you also need to decrement orig_qual_indirect, which counts the 
number of levels of array type derivation from orig_qual_type.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Marek Polacek]

On Thu, Dec 03, 2015 at 06:11:42PM +0000, Joseph Myers wrote:
> On Thu, 3 Dec 2015, Marek Polacek wrote:
> 
> > This ought to fix the fallout from PR c/68162 fix.  Here the problem is that
> > grokdeclarator created a wrong type for PARM_DECL "p".  It created this decl
> > with type "const int[<unknown>] *" while it should be "const int *".
> > 
> > I think the problem is that we weren't using TREE_TYPE on orig_qual_type and
> > thus c_build_qualified_type and subsequent c_build_pointer_type might create
> > a bogus type.  So when we're transfering const-ness of an array into that of
> > type pointed to, use TREE_TYPE not only of "type", but even of the orig qual
> > type.
> 
> I think you also need to decrement orig_qual_indirect, which counts the 
> number of levels of array type derivation from orig_qual_type.

Thus:

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-12-03  Marek Polacek  <polacek@redhat.com>

	PR c/68668
	* c-decl.c (grokdeclarator): When creating a PARM_DECL of ARRAY_TYPE,
	use TREE_TYPE of orig_qual_type.  Decrement ORIG_QUAL_INDIRECT.

	* gcc.dg/pr68668.c: New test.

diff --git gcc/c/c-decl.c gcc/c/c-decl.c
index 9ad8219..25bd1e0 100644
--- gcc/c/c-decl.c
+++ gcc/c/c-decl.c
@@ -6417,6 +6417,11 @@ grokdeclarator (const struct c_declarator *declarator,
 	  {
 	    /* Transfer const-ness of array into that of type pointed to.  */
 	    type = TREE_TYPE (type);
+	    if (orig_qual_type != NULL_TREE)
+	      {
+		orig_qual_type = TREE_TYPE (orig_qual_type);
+		orig_qual_indirect--;
+	      }
 	    if (type_quals)
 	      type = c_build_qualified_type (type, type_quals, orig_qual_type,
 					     orig_qual_indirect);
diff --git gcc/testsuite/gcc.dg/pr68668.c gcc/testsuite/gcc.dg/pr68668.c
index e69de29..d144fb6 100644
--- gcc/testsuite/gcc.dg/pr68668.c
+++ gcc/testsuite/gcc.dg/pr68668.c
@@ -0,0 +1,10 @@
+/* PR c/68668 */
+/* { dg-do compile } */
+
+typedef const int T[];
+
+int
+fn1 (T p)
+{
+  return p[0];
+}

	Marek

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Joseph Myers]

On Thu, 3 Dec 2015, Marek Polacek wrote:

> > I think you also need to decrement orig_qual_indirect, which counts the 
> > number of levels of array type derivation from orig_qual_type.
> 
> Thus:
> 
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
> 
> 2015-12-03  Marek Polacek  <polacek@redhat.com>
> 
> 	PR c/68668
> 	* c-decl.c (grokdeclarator): When creating a PARM_DECL of ARRAY_TYPE,
> 	use TREE_TYPE of orig_qual_type.  Decrement ORIG_QUAL_INDIRECT.

On further consideration:

Removing one level of array type derivation from type means it is one 
fewer levels indirect from the original version of orig_qual_type.  So I 
think you should actually decrement orig_qual_indirect without changing 
orig_qual_type.  But, if orig_qual_indirect is indirect, in that case you 
may get better results from changing orig_qual_type without decrementing 
orig_qual_indirect.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Marek Polacek]

On Thu, Dec 03, 2015 at 09:40:29PM +0000, Joseph Myers wrote:
> On Thu, 3 Dec 2015, Marek Polacek wrote:
> 
> > > I think you also need to decrement orig_qual_indirect, which counts the 
> > > number of levels of array type derivation from orig_qual_type.
> > 
> > Thus:
> > 
> > Bootstrapped/regtested on x86_64-linux, ok for trunk?
> > 
> > 2015-12-03  Marek Polacek  <polacek@redhat.com>
> > 
> > 	PR c/68668
> > 	* c-decl.c (grokdeclarator): When creating a PARM_DECL of ARRAY_TYPE,
> > 	use TREE_TYPE of orig_qual_type.  Decrement ORIG_QUAL_INDIRECT.
> 
> On further consideration:
> 
> Removing one level of array type derivation from type means it is one 
> fewer levels indirect from the original version of orig_qual_type.  So I 
> think you should actually decrement orig_qual_indirect without changing 
> orig_qual_type.  But, if orig_qual_indirect is indirect, in that case you 
> may get better results from changing orig_qual_type without decrementing 
> orig_qual_indirect.

I think I don't quite understand the last part of this, but here's the patch
anyway.  Having added more testcases, I noticed that my last version indeed
wasn't really correct -- thanks for spotting that.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-12-07  Marek Polacek  <polacek@redhat.com>

	PR c/68668
	* c-decl.c (grokdeclarator): If ORIG_QUAL_INDIRECT is indirect, use
	TREE_TYPE of ORIG_QUAL_TYPE, otherwise decrement ORIG_QUAL_INDIRECT.

	* gcc.dg/pr68668.c: New test.

diff --git gcc/c/c-decl.c gcc/c/c-decl.c
index 9ad8219..6a85514 100644
--- gcc/c/c-decl.c
+++ gcc/c/c-decl.c
@@ -6417,6 +6417,13 @@ grokdeclarator (const struct c_declarator *declarator,
 	  {
 	    /* Transfer const-ness of array into that of type pointed to.  */
 	    type = TREE_TYPE (type);
+	    if (orig_qual_type != NULL_TREE)
+	      {
+		if (orig_qual_indirect != 0)
+		  orig_qual_type = TREE_TYPE (orig_qual_type);
+		else
+		  orig_qual_indirect--;
+	      }
 	    if (type_quals)
 	      type = c_build_qualified_type (type, type_quals, orig_qual_type,
 					     orig_qual_indirect);
diff --git gcc/testsuite/gcc.dg/pr68668.c gcc/testsuite/gcc.dg/pr68668.c
index e69de29..d013aa9 100644
--- gcc/testsuite/gcc.dg/pr68668.c
+++ gcc/testsuite/gcc.dg/pr68668.c
@@ -0,0 +1,53 @@
+/* PR c/68668 */
+/* { dg-do compile } */
+
+typedef const int T[];
+typedef const int U[1];
+
+int
+fn1 (T p)
+{
+  return p[0];
+}
+
+int
+fn2 (U p[2])
+{
+  return p[0][0];
+}
+
+int
+fn3 (U p[2][3])
+{
+  return p[0][0][0];
+}
+
+int
+fn4 (U *p)
+{
+  return p[0][0];
+}
+
+int
+fn5 (U (*p)[1])
+{
+  return (*p)[0][0];
+}
+
+int
+fn6 (U (*p)[1][2])
+{
+  return (*p)[0][0][0];
+}
+
+int
+fn7 (U **p)
+{
+  return p[0][0][0];
+}
+
+int
+fn8 (U (**p)[1])
+{
+  return (*p)[0][0][0];
+}

	Marek

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Joseph Myers]

On Mon, 7 Dec 2015, Marek Polacek wrote:

> +		if (orig_qual_indirect != 0)
> +		  orig_qual_type = TREE_TYPE (orig_qual_type);
> +		else
> +		  orig_qual_indirect--;

For optimal results for debug info, I think that should be == 0 (i.e. 
preserve orig_qual_type, which may be a typedef, if possible - if the 
parameter is an array of orig_qual_type), rather than != 0.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Marek Polacek]

On Mon, Dec 07, 2015 at 04:05:11PM +0000, Joseph Myers wrote:
> On Mon, 7 Dec 2015, Marek Polacek wrote:
> 
> > +		if (orig_qual_indirect != 0)
> > +		  orig_qual_type = TREE_TYPE (orig_qual_type);
> > +		else
> > +		  orig_qual_indirect--;
> 
> For optimal results for debug info, I think that should be == 0 (i.e. 
> preserve orig_qual_type, which may be a typedef, if possible - if the 
> parameter is an array of orig_qual_type), rather than != 0.

I'm confused now.  If orig_qual_indirect != 0, it is indirect, right?
Earlier you said:
"if orig_qual_indirect is indirect, in that case you 
may get better results from changing orig_qual_type without decrementing 
orig_qual_indirect."
Isn't that something else than this version with == 0?

Anyway, here's the version with == 0.  Thanks,

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-12-07  Marek Polacek  <polacek@redhat.com>

	PR c/68668
	* c-decl.c (grokdeclarator): If ORIG_QUAL_INDIRECT is indirect, use
	TREE_TYPE of ORIG_QUAL_TYPE, otherwise decrement ORIG_QUAL_INDIRECT.

	* gcc.dg/pr68668.c: New test.

diff --git gcc/c/c-decl.c gcc/c/c-decl.c
index 9ad8219..6a85514 100644
--- gcc/c/c-decl.c
+++ gcc/c/c-decl.c
@@ -6417,6 +6417,13 @@ grokdeclarator (const struct c_declarator *declarator,
 	  {
 	    /* Transfer const-ness of array into that of type pointed to.  */
 	    type = TREE_TYPE (type);
+	    if (orig_qual_type != NULL_TREE)
+	      {
+		if (orig_qual_indirect == 0)
+		  orig_qual_type = TREE_TYPE (orig_qual_type);
+		else
+		  orig_qual_indirect--;
+	      }
 	    if (type_quals)
 	      type = c_build_qualified_type (type, type_quals, orig_qual_type,
 					     orig_qual_indirect);
diff --git gcc/testsuite/gcc.dg/pr68668.c gcc/testsuite/gcc.dg/pr68668.c
index e69de29..d013aa9 100644
--- gcc/testsuite/gcc.dg/pr68668.c
+++ gcc/testsuite/gcc.dg/pr68668.c
@@ -0,0 +1,53 @@
+/* PR c/68668 */
+/* { dg-do compile } */
+
+typedef const int T[];
+typedef const int U[1];
+
+int
+fn1 (T p)
+{
+  return p[0];
+}
+
+int
+fn2 (U p[2])
+{
+  return p[0][0];
+}
+
+int
+fn3 (U p[2][3])
+{
+  return p[0][0][0];
+}
+
+int
+fn4 (U *p)
+{
+  return p[0][0];
+}
+
+int
+fn5 (U (*p)[1])
+{
+  return (*p)[0][0];
+}
+
+int
+fn6 (U (*p)[1][2])
+{
+  return (*p)[0][0][0];
+}
+
+int
+fn7 (U **p)
+{
+  return p[0][0][0];
+}
+
+int
+fn8 (U (**p)[1])
+{
+  return (*p)[0][0][0];
+}

	Marek

Re: C PATCH for c/68668 (grokdeclarator and wrong type of PARM_DECL)

[Joseph Myers]

On Mon, 7 Dec 2015, Marek Polacek wrote:

> Anyway, here's the version with == 0.  Thanks,
> 
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
> 
> 2015-12-07  Marek Polacek  <polacek@redhat.com>
> 
> 	PR c/68668
> 	* c-decl.c (grokdeclarator): If ORIG_QUAL_INDIRECT is indirect, use
> 	TREE_TYPE of ORIG_QUAL_TYPE, otherwise decrement ORIG_QUAL_INDIRECT.
> 
> 	* gcc.dg/pr68668.c: New test.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com