[PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jakub Jelinek]

Hi!

cplus_decl_attributes can be called with attributes equal to
error_mark_node, there are some spots in the function that test
it or decl_attributes it calls starts with:
  if (TREE_TYPE (*node) == error_mark_node || attributes == error_mark_node)
    return NULL_TREE;
But the recent PR104245 change broke this when processing_template_decl
is true.

This fixes it and also fixes an OpenMP problem with such attributes.

Ok for trunk if it passes bootstrap/regtest?

2022-03-25  Jakub Jelinek  <jakub@redhat.com>

	PR c++/104668
	* decl2.cc (splice_template_attributes): Return NULL if *p is
	error_mark_node.
	(cplus_decl_attributes): Don't chain on OpenMP attributes if
	attributes is error_mark_node.

	* g++.dg/cpp0x/pr104668.C: New test.

--- gcc/cp/decl2.cc.jj	2022-03-09 09:09:55.415843331 +0100
+++ gcc/cp/decl2.cc	2022-03-25 17:17:27.769036749 +0100
@@ -1336,7 +1336,7 @@ splice_template_attributes (tree *attr_p
   tree late_attrs = NULL_TREE;
   tree *q = &late_attrs;
 
-  if (!p)
+  if (!p || *p == error_mark_node)
     return NULL_TREE;
 
   for (; *p; )
@@ -1644,6 +1644,8 @@ cplus_decl_attributes (tree *decl, tree
 	  && DECL_CLASS_SCOPE_P (*decl))
 	error ("%q+D static data member inside of declare target directive",
 	       *decl);
+      else if (attributes == error_mark_node)
+	;
       else if (VAR_P (*decl)
 	       && (processing_template_decl
 		   || !cp_omp_mappable_type (TREE_TYPE (*decl))))
--- gcc/testsuite/g++.dg/cpp0x/pr104668.C.jj	2022-03-25 17:25:42.280068058 +0100
+++ gcc/testsuite/g++.dg/cpp0x/pr104668.C	2022-03-25 17:24:44.862881444 +0100
@@ -0,0 +1,13 @@
+// PR c++/104668
+// { dg-do compile { target c++11 } }
+// { dg-excess-errors "" }
+
+template <class... Ts>
+void sink(Ts...);
+template <class... Ts>
+void f(Ts...) {
+  sink([] { struct alignas:Ts) S {}; }...); }
+}
+int main() {
+  f(0);
+}

	Jakub

Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jason Merrill]

On 3/25/22 12:34, Jakub Jelinek wrote:
> Hi!
> 
> cplus_decl_attributes can be called with attributes equal to
> error_mark_node, there are some spots in the function that test
> it or decl_attributes it calls starts with:
>    if (TREE_TYPE (*node) == error_mark_node || attributes == error_mark_node)
>      return NULL_TREE;
> But the recent PR104245 change broke this when processing_template_decl
> is true.
> 
> This fixes it and also fixes an OpenMP problem with such attributes.
> 
> Ok for trunk if it passes bootstrap/regtest?
> 
> 2022-03-25  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR c++/104668
> 	* decl2.cc (splice_template_attributes): Return NULL if *p is
> 	error_mark_node.
> 	(cplus_decl_attributes): Don't chain on OpenMP attributes if
> 	attributes is error_mark_node.
> 
> 	* g++.dg/cpp0x/pr104668.C: New test.
> 
> --- gcc/cp/decl2.cc.jj	2022-03-09 09:09:55.415843331 +0100
> +++ gcc/cp/decl2.cc	2022-03-25 17:17:27.769036749 +0100
> @@ -1336,7 +1336,7 @@ splice_template_attributes (tree *attr_p
>     tree late_attrs = NULL_TREE;
>     tree *q = &late_attrs;
>   
> -  if (!p)
> +  if (!p || *p == error_mark_node)
>       return NULL_TREE;
>   
>     for (; *p; )
> @@ -1644,6 +1644,8 @@ cplus_decl_attributes (tree *decl, tree
>   	  && DECL_CLASS_SCOPE_P (*decl))
>   	error ("%q+D static data member inside of declare target directive",
>   	       *decl);
> +      else if (attributes == error_mark_node)
> +	;

Why not check at the beginning of the function?

>         else if (VAR_P (*decl)
>   	       && (processing_template_decl
>   		   || !cp_omp_mappable_type (TREE_TYPE (*decl))))
> --- gcc/testsuite/g++.dg/cpp0x/pr104668.C.jj	2022-03-25 17:25:42.280068058 +0100
> +++ gcc/testsuite/g++.dg/cpp0x/pr104668.C	2022-03-25 17:24:44.862881444 +0100
> @@ -0,0 +1,13 @@
> +// PR c++/104668
> +// { dg-do compile { target c++11 } }
> +// { dg-excess-errors "" }
> +
> +template <class... Ts>
> +void sink(Ts...);
> +template <class... Ts>
> +void f(Ts...) {
> +  sink([] { struct alignas:Ts) S {}; }...); }
> +}
> +int main() {
> +  f(0);
> +}
> 
> 	Jakub
> 

Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jakub Jelinek]

On Fri, Mar 25, 2022 at 07:48:33PM +0100, Jakub Jelinek wrote:
> We then wouldn't propagate TREE_DEPRECATED/TREE_UNAVAILABLE from templates
> to their instantiations and wouldn't diagnose static data members in OpenMP
> declare target.
> But perhaps that is fine, with error_mark_attribute it is error recovery
> anyway.
> Can the splice_template_attributes change stay too, or just
> change cplus_decl_attributes (add it to the early out, remove from other
> spots in the function)?

So like (or without the first hunk):

2022-03-25  Jakub Jelinek  <jakub@redhat.com>

	PR c++/104668
	* decl2.cc (splice_template_attributes): Return NULL if *p is
	error_mark_node.
	(cplus_decl_attributes): Return early if attributes is
	error_mark_node.  Don't check that later.

	* g++.dg/cpp0x/pr104668.C: New test.

--- gcc/cp/decl2.cc.jj	2022-03-09 09:09:55.415843331 +0100
+++ gcc/cp/decl2.cc	2022-03-25 19:54:12.454749089 +0100
@@ -1336,7 +1336,7 @@ splice_template_attributes (tree *attr_p
   tree late_attrs = NULL_TREE;
   tree *q = &late_attrs;
 
-  if (!p)
+  if (!p || *p == error_mark_node)
     return NULL_TREE;
 
   for (; *p; )
@@ -1631,7 +1631,7 @@ void
 cplus_decl_attributes (tree *decl, tree attributes, int flags)
 {
   if (*decl == NULL_TREE || *decl == void_type_node
-      || *decl == error_mark_node)
+      || *decl == error_mark_node || attributes == error_mark_node)
     return;
 
   /* Add implicit "omp declare target" attribute if requested.  */
@@ -1668,7 +1668,7 @@ cplus_decl_attributes (tree *decl, tree
 
   cp_check_const_attributes (attributes);
 
-  if ((flag_openmp || flag_openmp_simd) && attributes != error_mark_node)
+  if (flag_openmp || flag_openmp_simd)
     {
       bool diagnosed = false;
       for (tree *pa = &attributes; *pa; )
--- gcc/testsuite/g++.dg/cpp0x/pr104668.C.jj	2022-03-25 17:25:42.280068058 +0100
+++ gcc/testsuite/g++.dg/cpp0x/pr104668.C	2022-03-25 17:24:44.862881444 +0100
@@ -0,0 +1,13 @@
+// PR c++/104668
+// { dg-do compile { target c++11 } }
+// { dg-excess-errors "" }
+
+template <class... Ts>
+void sink(Ts...);
+template <class... Ts>
+void f(Ts...) {
+  sink([] { struct alignas:Ts) S {}; }...); }
+}
+int main() {
+  f(0);
+}


	Jakub

Ping Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jakub Jelinek]

On Fri, Mar 25, 2022 at 07:55:56PM +0100, Jakub Jelinek via Gcc-patches wrote:
> On Fri, Mar 25, 2022 at 07:48:33PM +0100, Jakub Jelinek wrote:
> > We then wouldn't propagate TREE_DEPRECATED/TREE_UNAVAILABLE from templates
> > to their instantiations and wouldn't diagnose static data members in OpenMP
> > declare target.
> > But perhaps that is fine, with error_mark_attribute it is error recovery
> > anyway.
> > Can the splice_template_attributes change stay too, or just
> > change cplus_decl_attributes (add it to the early out, remove from other
> > spots in the function)?
> 
> So like (or without the first hunk):
> 
> 2022-03-25  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR c++/104668
> 	* decl2.cc (splice_template_attributes): Return NULL if *p is
> 	error_mark_node.
> 	(cplus_decl_attributes): Return early if attributes is
> 	error_mark_node.  Don't check that later.
> 
> 	* g++.dg/cpp0x/pr104668.C: New test.

I'd like to ping this patch.  Bootstrapped/regtested successfully on
{powerpc64le,x86_64,i686}-linux.

> --- gcc/cp/decl2.cc.jj	2022-03-09 09:09:55.415843331 +0100
> +++ gcc/cp/decl2.cc	2022-03-25 19:54:12.454749089 +0100
> @@ -1336,7 +1336,7 @@ splice_template_attributes (tree *attr_p
>    tree late_attrs = NULL_TREE;
>    tree *q = &late_attrs;
>  
> -  if (!p)
> +  if (!p || *p == error_mark_node)
>      return NULL_TREE;
>  
>    for (; *p; )
> @@ -1631,7 +1631,7 @@ void
>  cplus_decl_attributes (tree *decl, tree attributes, int flags)
>  {
>    if (*decl == NULL_TREE || *decl == void_type_node
> -      || *decl == error_mark_node)
> +      || *decl == error_mark_node || attributes == error_mark_node)
>      return;
>  
>    /* Add implicit "omp declare target" attribute if requested.  */
> @@ -1668,7 +1668,7 @@ cplus_decl_attributes (tree *decl, tree
>  
>    cp_check_const_attributes (attributes);
>  
> -  if ((flag_openmp || flag_openmp_simd) && attributes != error_mark_node)
> +  if (flag_openmp || flag_openmp_simd)
>      {
>        bool diagnosed = false;
>        for (tree *pa = &attributes; *pa; )
> --- gcc/testsuite/g++.dg/cpp0x/pr104668.C.jj	2022-03-25 17:25:42.280068058 +0100
> +++ gcc/testsuite/g++.dg/cpp0x/pr104668.C	2022-03-25 17:24:44.862881444 +0100
> @@ -0,0 +1,13 @@
> +// PR c++/104668
> +// { dg-do compile { target c++11 } }
> +// { dg-excess-errors "" }
> +
> +template <class... Ts>
> +void sink(Ts...);
> +template <class... Ts>
> +void f(Ts...) {
> +  sink([] { struct alignas:Ts) S {}; }...); }
> +}
> +int main() {
> +  f(0);
> +}

	Jakub

Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jason Merrill]

On 3/25/22 14:08, Jason Merrill wrote:
> On 3/25/22 12:34, Jakub Jelinek wrote:
>> Hi!
>>
>> cplus_decl_attributes can be called with attributes equal to
>> error_mark_node, there are some spots in the function that test
>> it or decl_attributes it calls starts with:
>>    if (TREE_TYPE (*node) == error_mark_node || attributes == 
>> error_mark_node)
>>      return NULL_TREE;
>> But the recent PR104245 change broke this when processing_template_decl
>> is true.
>>
>> This fixes it and also fixes an OpenMP problem with such attributes.
>>
>> Ok for trunk if it passes bootstrap/regtest?
>>
>> 2022-03-25  Jakub Jelinek  <jakub@redhat.com>
>>
>>     PR c++/104668
>>     * decl2.cc (splice_template_attributes): Return NULL if *p is
>>     error_mark_node.
>>     (cplus_decl_attributes): Don't chain on OpenMP attributes if
>>     attributes is error_mark_node.
>>
>>     * g++.dg/cpp0x/pr104668.C: New test.
>>
>> --- gcc/cp/decl2.cc.jj    2022-03-09 09:09:55.415843331 +0100
>> +++ gcc/cp/decl2.cc    2022-03-25 17:17:27.769036749 +0100
>> @@ -1336,7 +1336,7 @@ splice_template_attributes (tree *attr_p
>>     tree late_attrs = NULL_TREE;
>>     tree *q = &late_attrs;
>> -  if (!p)
>> +  if (!p || *p == error_mark_node)
>>       return NULL_TREE;
>>     for (; *p; )
>> @@ -1644,6 +1644,8 @@ cplus_decl_attributes (tree *decl, tree
>>         && DECL_CLASS_SCOPE_P (*decl))
>>       error ("%q+D static data member inside of declare target 
>> directive",
>>              *decl);
>> +      else if (attributes == error_mark_node)
>> +    ;
> 
> Why not check at the beginning of the function?

You just pinged this patch, but I haven't seen a response to this question.

>>         else if (VAR_P (*decl)
>>              && (processing_template_decl
>>              || !cp_omp_mappable_type (TREE_TYPE (*decl))))
>> --- gcc/testsuite/g++.dg/cpp0x/pr104668.C.jj    2022-03-25 
>> 17:25:42.280068058 +0100
>> +++ gcc/testsuite/g++.dg/cpp0x/pr104668.C    2022-03-25 
>> 17:24:44.862881444 +0100
>> @@ -0,0 +1,13 @@
>> +// PR c++/104668
>> +// { dg-do compile { target c++11 } }
>> +// { dg-excess-errors "" }
>> +
>> +template <class... Ts>
>> +void sink(Ts...);
>> +template <class... Ts>
>> +void f(Ts...) {
>> +  sink([] { struct alignas:Ts) S {}; }...); }
>> +}
>> +int main() {
>> +  f(0);
>> +}
>>
>>     Jakub
>>
> 

Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jakub Jelinek]

On Wed, Apr 06, 2022 at 11:18:32AM -0400, Jason Merrill wrote:
> > Why not check at the beginning of the function?
> 
> You just pinged this patch, but I haven't seen a response to this question.

I thought the
https://gcc.gnu.org/pipermail/gcc-patches/2022-March/592368.html
is the response to that.
In my =sent I have also a mail which just replied to your
> Why not check at the beginning of the function?
but didn't contain any patch, apparently that wasn't sent due to
some outgoing mail sending issues.
But the above mail is a reply to that mail citing that
reply and including a patch.

	Jakub

Re: [PATCH] c++: Fix up ICE when cplus_decl_attributes is called with error_mark_node attributes [PR104668]

[Jason Merrill]

On 4/6/22 11:26, Jakub Jelinek wrote:
> On Wed, Apr 06, 2022 at 11:18:32AM -0400, Jason Merrill wrote:
>>> Why not check at the beginning of the function?
>>
>> You just pinged this patch, but I haven't seen a response to this question.
> 
> I thought the
> https://gcc.gnu.org/pipermail/gcc-patches/2022-March/592368.html
> is the response to that.

Weird, AFAICT I never got that message.  But yes, that patch is OK.

> In my =sent I have also a mail which just replied to your
>> Why not check at the beginning of the function?
> but didn't contain any patch, apparently that wasn't sent due to
> some outgoing mail sending issues.
> But the above mail is a reply to that mail citing that
> reply and including a patch.