Re: debug/6436: dwarf2out ICE with typedef using attributes

Re: debug/6436: dwarf2out ICE with typedef using attributes

[jakub]

Old Synopsis: 
New Synopsis: dwarf2out ICE with typedef using attributes

State-Changed-From-To: open->analyzed
State-Changed-By: jakub
State-Changed-When: Mon Apr 29 05:31:49 2002
State-Changed-Why:
    Simplified into:
    --- gcc/testsuite/g++.dg/debug/debug6.C.jj      Thu Aug 30 22:30:55 2001
    +++ gcc/testsuite/g++.dg/debug/debug6.C Mon Apr 29 14:41:05 2002
    @@ -0,0 +1,17 @@
    +// PR debug/6436
    +// { dg-do compile }
    +
    +typedef struct
    +{
    +  unsigned int a0, a1;
    +} A __attribute__ ((aligned(8)));
    +
    +typedef struct
    +{
    +  A a;
    +} B;
    +
    +struct C
    +{
    +  B *bp;
    +};
    
    Regression from 2.9x (even with -gdwarf-2).

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6436

Re: debug/6436: dwarf2out ICE with typedef using attributes

[jason]

Synopsis: dwarf2out ICE with typedef using attributes

State-Changed-From-To: analyzed->closed
State-Changed-By: jason
State-Changed-When: Wed May  1 03:19:57 2002
State-Changed-Why:
    Fixed.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6436

Re: debug/6436: dwarf2out ICE with typedef using attributes

[Jason Merrill]

The following reply was made to PR debug/6436; it has been noted by GNATS.

From: Jason Merrill <jason@redhat.com>
To: Daniel Berlin <dberlin@dberlin.org>
Cc: jakub@gcc.gnu.org, <gcc-bugs@gcc.gnu.org>, <meissner@suse.de>,
	<gcc-gnats@gcc.gnu.org>, <gcc-patches@gcc.gnu.org>,
	Mark Mitchell <mark@codesourcery.com>,
	Richard Sandiford <rsandifo@redhat.com>
Subject: Re: debug/6436: dwarf2out ICE with typedef using attributes
Date: Tue, 30 Apr 2002 18:48:33 +0100

 --=-=-=
 
 >>>>> "Jason" == Jason Merrill <jason@redhat.com> writes:
 
 > One fix would be to not give the name 'A' to the anonymous struct.
 
 Thus.  I've added Jakub's testcase to the testsuite as
 g++.dg/debug/typedef1.C.
 
 Booted and tested i686-pc-linux-gnu, applied trunk and 3.1.
 
 2002-04-30  Jason Merrill  <jason@redhat.com>
 
 	* decl.c (grokdeclarator): Don't override TYPE_NAME of an
 	anonymous class with a typedef if there are attributes.
 
 
 --=-=-=
 Content-Type: text/x-patch
 Content-Disposition: inline
 
 *** decl.c.~1~	Tue Apr 30 14:06:13 2002
 --- decl.c	Tue Apr 30 18:21:09 2002
 *************** grokdeclarator (declarator, declspecs, d
 *** 11275,11280 ****
 --- 11275,11282 ----
   	  && TYPE_NAME (type)
   	  && TREE_CODE (TYPE_NAME (type)) == TYPE_DECL
   	  && TYPE_ANONYMOUS_P (type)
 + 	  /* Don't do this if there are attributes.  */
 + 	  && (!attrlist || !*attrlist)
   	  && cp_type_quals (type) == TYPE_UNQUALIFIED)
   	{
   	  tree oldname = TYPE_NAME (type);
 
 --=-=-=--

Re: debug/6436: dwarf2out ICE with typedef using attributes

[Jason Merrill]

The following reply was made to PR debug/6436; it has been noted by GNATS.

From: Jason Merrill <jason@redhat.com>
To: Daniel Berlin <dberlin@dberlin.org>
Cc: jakub@gcc.gnu.org, <gcc-bugs@gcc.gnu.org>, <meissner@suse.de>,
	<gcc-gnats@gcc.gnu.org>, <gcc-patches@gcc.gnu.org>,
	Mark Mitchell <mark@codesourcery.com>,
	Richard Sandiford <rsandifo@redhat.com>
Subject: Re: debug/6436: dwarf2out ICE with typedef using attributes
Date: Tue, 30 Apr 2002 11:01:51 +0100

 >>>>> "Daniel" == Daniel Berlin <dberlin@dberlin.org> writes:
 
 > For C++, we output two *named* types first, ._0, which is the A *without* 
 > alignment, then ._1, which is the A *with* alignment.
 
 Actually, ._1 is B.  The C++ standard says
 
   If the typedef declaration defines an unnamed  class  (or  enum),  the
   first  typedef-name  declared by the declaration to be that class type
   (or enum type) is used to denote the class type  (or  enum  type)  for
   linkage purposes only.
 
 So we set TYPE_NAME for the type to be the typedef, rather than the funny
 internal name.  For B this is fine, because the type of the name B is the
 struct itself.  For A we have a problem: the type of the name A is an
 attribute-variant of the anonymous struct, but we set the struct's name
 anyway.  So
 
   TREE_TYPE (TYPE_NAME (type)) != type
 
 but
 
   DECL_ORIGINAL_TYPE (TYPE_NAME (type)) == type
 
 One fix would be to not give the name 'A' to the anonymous struct.
 However, that would mean that we couldn't use the type in a mangled
 context, and I suspect that there is code out there that relies on this
 sort of pattern, though I suppose it's easy enough to fix affected code by
 adding an "A" after the "struct" tag.
 
 Jason

Re: debug/6436: dwarf2out ICE with typedef using attributes

[Daniel Berlin]

The following reply was made to PR debug/6436; it has been noted by GNATS.

From: Daniel Berlin <dberlin@dberlin.org>
To: Jason Merrill <jason@redhat.com>
Cc: jakub@gcc.gnu.org, <gcc-bugs@gcc.gnu.org>, <meissner@suse.de>,
	<gcc-gnats@gcc.gnu.org>, <gcc-patches@gcc.gnu.org>,
	Mark Mitchell <mark@codesourcery.com>,
	Richard Sandiford <rsandifo@redhat.com>
Subject: Re: debug/6436: dwarf2out ICE with typedef using attributes
Date: Mon, 29 Apr 2002 14:32:14 -0400 (EDT)

 On Mon, 29 Apr 2002, Jason Merrill wrote:
 
 > This fix seems reasonable; I'll apply it after testing.  I'm a bit
 > uncomfortable with the typedef pointing to its own type, though; what do
 > other people think the effect of Jakub's testcase should be?
 > 
 > Richard Sandiford's recent patch tries to deal with this sort of situation
 > by separating the typedef type and underlying struct type; why isn't that
 > working here?
 > 
 
 Here's a kicker, probably related:  The exact code in question works 
 perfectly in C.  IE run jakub's test under cc1, and it does fine.
 
 In fact, I was about to email Jakub saying i couldn't reproduce, wwhne i 
 noticed the original testcase was in C++. So i tried cc1plus, and boom.
 
 
 Analyzing what goes right in C:
 
 We start the same path, getting the member_declared_type of a, seeing it's 
 A, and going to output a type die for A *with* alignment.
 
 It gets the main variant, which, just like C++, is A *without* alignment.
 ...
 
 Okay, well, by that point, the A *without* alignment is already written 
 out, so it returns right there.
 
 Which is the difference and why C works.
 SO let me back up.
 
 The first type we output for C is an unnamed type that is A *without* 
 alignment.
 I.E. TYPE_NAME (type) == NULL
 It's output from finish_struct, which calls rest_of_type_compilation, 
 which calls dwarf2out_decl (TYPE_STUB_DECL (type)).
 
 Then we output unsigned int, then another unnamed type that is A with 
 alignment.
 
 Then we get to output A, whose main variant is the *unnamed* type we wrote 
 out before.  Thus, TYPE_NAME (type) == NULL, and we go about our merry 
 way.
 So it works fine.
 
 
 For C++, we output two *named* types first, ._0, which is the A *without* 
 alignment, then ._1, which is the A *with* alignment.
 I.E. TYPE_NAME (type) != NULL on either of them.
 
 Eventually we go to output A, whose main variant is not either of the 
 named type, but another type named "A" without the alignment.
 
 This is what starts the dominoes falling.
 
 If the main variant of A was ._0, we should be fine.
 Or, if we output the second "A" type that seemed to materialize out of 
 nowhere, before trying to output the real "A", we'd be fine.
 
 My gut feeling is the main variant of A with alignment should be ._0, not 
 this other type named "A" without alignment that seemed to materialize out 
 of nowhere, and was never output.
 
 But this is just because it "feels" right, and what C does.
 I may be completely wrong.
 
 > 2002-04-29  Jason Merrill  <jason@redhat.com>
 > 
 > 	* dwarf2out.c (gen_type_die): Don't recurse on a self-referential
 > 	typedef.
 > 
 > 
 

Re: debug/6436: dwarf2out ICE with typedef using attributes

[Jason Merrill]

The following reply was made to PR debug/6436; it has been noted by GNATS.

From: Jason Merrill <jason@redhat.com>
To: Daniel Berlin <dberlin@dberlin.org>
Cc: jakub@gcc.gnu.org, <gcc-bugs@gcc.gnu.org>, <meissner@suse.de>,
	<gcc-gnats@gcc.gnu.org>, gcc-patches@gcc.gnu.org,
	Mark Mitchell <mark@codesourcery.com>,
	Richard Sandiford <rsandifo@redhat.com>
Subject: Re: debug/6436: dwarf2out ICE with typedef using attributes
Date: Mon, 29 Apr 2002 18:47:02 +0100

 --=-=-=
 
 This fix seems reasonable; I'll apply it after testing.  I'm a bit
 uncomfortable with the typedef pointing to its own type, though; what do
 other people think the effect of Jakub's testcase should be?
 
 Richard Sandiford's recent patch tries to deal with this sort of situation
 by separating the typedef type and underlying struct type; why isn't that
 working here?
 
 2002-04-29  Jason Merrill  <jason@redhat.com>
 
 	* dwarf2out.c (gen_type_die): Don't recurse on a self-referential
 	typedef.
 
 
 --=-=-=
 Content-Type: text/x-patch
 Content-Disposition: inline
 
 *** dwarf2out.c.~1~	Mon Apr 29 18:39:20 2002
 --- dwarf2out.c	Mon Apr 29 18:15:37 2002
 *************** gen_type_die (type, context_die)
 *** 11068,11074 ****
       return;
   
     if (TYPE_NAME (type) && TREE_CODE (TYPE_NAME (type)) == TYPE_DECL
 !       && DECL_ORIGINAL_TYPE (TYPE_NAME (type)))
       {
         TREE_ASM_WRITTEN (type) = 1;
         gen_decl_die (TYPE_NAME (type), context_die);
 --- 11068,11075 ----
       return;
   
     if (TYPE_NAME (type) && TREE_CODE (TYPE_NAME (type)) == TYPE_DECL
 !       && DECL_ORIGINAL_TYPE (TYPE_NAME (type))
 !       && DECL_ORIGINAL_TYPE (TYPE_NAME (type)) != type)
       {
         TREE_ASM_WRITTEN (type) = 1;
         gen_decl_die (TYPE_NAME (type), context_die);
 
 --=-=-=--

Re: debug/6436: dwarf2out ICE with typedef using attributes

[Daniel Berlin]

The following reply was made to PR debug/6436; it has been noted by GNATS.

From: Daniel Berlin <dberlin@dberlin.org>
To: jakub@gcc.gnu.org, <gcc-bugs@gcc.gnu.org>, <gcc-prs@gcc.gnu.org>,
	<jason@redhat.com>, <meissner@suse.de>, <nobody@gcc.gnu.org>,
	<gcc-gnats@gcc.gnu.org>, <jason@redhat.com>
Cc:  
Subject: Re: debug/6436: dwarf2out ICE with typedef using attributes
Date: Mon, 29 Apr 2002 13:31:54 -0400 (EDT)

 This one is fun.  Not playing beach volley ball fun, more like 
 burning yourself on hot coals fun.
 Here's what happens:
 
 We go to output a type die for A *with* alignment, because it's the type 
 of a decl.
 
 In gen_type_die, we first get the main variant of A, which is an A 
 *without* the user set alignment.
 
 We then hit this test near the top:
 
 if (TYPE_NAME (decl) && TREE_CODE (TYPE_NAME (type)) == TYPE_DECL && 
 	DECL_ORIGINAL_TYPE (TYPE_NAME (type)))
 
 and it succeeds, causing us to do:
 
 {
 <type is now a without alignment, type_name (type) is a with alignment>
 	TREE_ASM_WRITTEN (type) = 1;
 	gen_decl_die (TYPE_NAME (type), context_die);
 	return;
 }
 
 So we mark the A *without* alignment as written, then proceed to try to 
 gen a decl die for the type_name of A, which is just like A, only *with* 
 the alignment. (you should already smell trouble brewing here, since this 
 whole thing started when we were trying to output a type die for A *with* 
 alignment, and surely, gen_decl_die will try to do this as well)
 
 is_redundant_typedef on it says no, so we go to gen a typedef die for the 
 A with alignment, to the A without alignment.
 
 In gen_typedef_die, first we mark the A with alignment type as written, 
 then proceed to get the original type, A *without* alignment, and ask 
 modified_type_die for a type die with the qualifiers in front of it, if 
 any.  
 In the portion of modified_type_die labeled "Handle C typedef types", we 
 pass the main test, but qualified_type != dtype because one is the one 
 with alignment, one is the one without ailgnment, thus, they are 
 different trees, and the pointer comparison fails.  Even though, going 
 by qualifications, they *are* the same type.
 
 After failing that, we fail the other tests, getting us to the else which 
 calls gen_type_die, with the type passed in (This is the A without 
 alignment type, once again).
 Which of course, sees it's written the type (since we set the written flag 
 on that one at the beginning of this godawful trip through hell), and 
 returns.
 
 The lookup, which should succeed, and would, had we actually ever written 
 the original type, instead of us taking this trip to the 9th circle, 
 fails, and we abort.
 
 To be more concise, here's what happens:
 
 gen_type_die (A with alignment)
         |
 get main variant, A *without* alignment, see it's got a type name, 
 therefore, so mark A *without* alignment as written, and call
 	|
 gen_decl_die (TYPE_NAME (A without alignment), which is A with alignment)
         |
 is_redundant == no, so
         |
 gen_typedef (A with alignment) 
 	|
 Mark A with alignment as written.
         |
 Notice A with alignment has an original type, get original type, A 
 *without* alignment.
 	|
 modified_type_die (A *without* alignment)
 	|
 fail test saying it's equal to existing qualified type, causing us to:
 	|
 gen_type_die (A *without* alignment)
 	|
 Sees it's marked as written, so it just returns.
 	|
 lookup fails, since we really haven't written it (we are in the process of 
 trying to write it, we will when we get back up to the top of the 
 recursion chain), and we abort.
 
 
 My head hurts.
 
 I'm not sure what the right thing to do in this case is, so i'm not sure i 
 can point out a stage and say "That's where it all went wrong".  But 
 someone who *does* understand what we should do, should be able to use 
 this analysis to fix the problem.
 
 --Dan
  yet. On 29 Apr 2002 jakub@gcc.gnu.org 
 wrote:
 
 > Old Synopsis: 
 > New Synopsis: dwarf2out ICE with typedef using attributes
 > 
 > State-Changed-From-To: open->analyzed
 > State-Changed-By: jakub
 > State-Changed-When: Mon Apr 29 05:31:49 2002
 > State-Changed-Why:
 >     Simplified into:
 >     --- gcc/testsuite/g++.dg/debug/debug6.C.jj      Thu Aug 30 22:30:55 2001
 >     +++ gcc/testsuite/g++.dg/debug/debug6.C Mon Apr 29 14:41:05 2002
 >     @@ -0,0 +1,17 @@
 >     +// PR debug/6436
 >     +// { dg-do compile }
 >     +
 >     +typedef struct
 >     +{
 >     +  unsigned int a0, a1;
 >     +} A __attribute__ ((aligned(8)));
 >     +
 >     +typedef struct
 >     +{
 >     +  A a;
 >     +} B;
 >     +
 >     +struct C
 >     +{
 >     +  B *bp;
 >     +};
 >     
 >     Regression from 2.9x (even with -gdwarf-2).
 > 
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6436
 > 
 

Re: debug/6436: dwarf2out ICE with typedef using attributes

[jason]

Synopsis: dwarf2out ICE with typedef using attributes

Responsible-Changed-From-To: unassigned->jason
Responsible-Changed-By: jason
Responsible-Changed-When: Mon Apr 29 10:09:02 2002
Responsible-Changed-Why:
    got it

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6436