public inbox for gcc-help@gcc.gnu.org
 help / color / mirror / Atom feed
* Incorrect replacement of TYPE_METHODS with TYPE_FIELDS causing segfault.
@ 2022-06-08  4:47 Zopolis0
  2022-06-08  7:02 ` Xi Ruoyao
  0 siblings, 1 reply; 3+ messages in thread
From: Zopolis0 @ 2022-06-08  4:47 UTC (permalink / raw)
  To: gcc-help

While compiling gcj on my mstermerge branch (
https://github.com/Zopolis4/gcj/tree/mstermerge) it breaks on, amongst
other things, an internal compiler error:
echo ../../../../gcj/libjava/classpath/lib/java/lang/ref/*.class >
java/lang/ref.list
echo ../../../../gcj/libjava/classpath/lib/java/lang/reflect/*.class >
java/lang/reflect.list
In file included from <built-in>:31:
../../../../gcj/libjava/java/lang/Object.java: internal compiler error:
tree check: expected function_decl, have field_decl in layout_class_method,
at java/class.cc:2627
In file included from <built-in>:63:
java/lang/Object.java: internal compiler error: tree check: expected
function_decl, have field_decl in layout_class_method, at java/class.cc:2627

I believe that this error is due to the following incorrect code at line
2589 at gcc/java/class.cc:
  for (tree method_decl = TYPE_FIELDS (this_class);

Before it was removed, this line was:
  for (method_decl = TYPE_METHODS (this_class);

But I have since modified it (
https://github.com/Zopolis4/gcj/commit/1f38bc896a704290ca0b742c60c60a88d5e1fb07#diff-bf49cdc948b20b9f25afd3fbb36922a10e979a55ff9600b3b51d0ec54175b752L2592)
in accordance with 5aaa8fb40681ee66282d73dab8c8eccbf5ee0518

Given that this was incorrect, what would be the correct way to replace
this instance of TYPE_METHODS? Were all my replacements of TYPE_METHODS in
gcc/java/class.cc wrong?

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: Incorrect replacement of TYPE_METHODS with TYPE_FIELDS causing segfault.
  2022-06-08  4:47 Incorrect replacement of TYPE_METHODS with TYPE_FIELDS causing segfault Zopolis0
@ 2022-06-08  7:02 ` Xi Ruoyao
  2022-06-08  8:13   ` Zopolis0
  0 siblings, 1 reply; 3+ messages in thread
From: Xi Ruoyao @ 2022-06-08  7:02 UTC (permalink / raw)
  To: Zopolis0, gcc-help

On Wed, 2022-06-08 at 14:47 +1000, Zopolis0 via Gcc-help wrote:

/* snip */

> java/lang/Object.java: internal compiler error: tree check: expected
> function_decl, have field_decl in layout_class_method, at java/class.cc:2627
> 
> I believe that this error is due to the following incorrect code at line
> 2589 at gcc/java/class.cc:
>   for (tree method_decl = TYPE_FIELDS (this_class);
> 
> Before it was removed, this line was:
>   for (method_decl = TYPE_METHODS (this_class);
> 
> But I have since modified it (
> https://github.com/Zopolis4/gcj/commit/1f38bc896a704290ca0b742c60c60a88d5e1fb07#diff-bf49cdc948b20b9f25afd3fbb36922a10e979a55ff9600b3b51d0ec54175b752L2592)
> in accordance with 5aaa8fb40681ee66282d73dab8c8eccbf5ee0518
> 
> Given that this was incorrect, what would be the correct way to replace
> this instance of TYPE_METHODS? Were all my replacements of TYPE_METHODS in
> gcc/java/class.cc wrong?

5aaa8fb is not a simple replacement.  For example:

   /* If there are user-defined methods, they are deemed non-trivial.  */
-  for (tmp = TYPE_METHODS (type); tmp; tmp = TREE_CHAIN (tmp))
-    if (!DECL_ARTIFICIAL (tmp))
+  for (tree fld = TYPE_FIELDS (type); fld; fld = DECL_CHAIN (fld))
+    if (TREE_CODE (TREE_TYPE (fld)) == METHOD_TYPE && !DECL_ARTIFICIAL (fld))
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
       return true;

I think you need to check if the field is really a method like this.
-- 
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: Incorrect replacement of TYPE_METHODS with TYPE_FIELDS causing segfault.
  2022-06-08  7:02 ` Xi Ruoyao
@ 2022-06-08  8:13   ` Zopolis0
  0 siblings, 0 replies; 3+ messages in thread
From: Zopolis0 @ 2022-06-08  8:13 UTC (permalink / raw)
  To: Xi Ruoyao; +Cc: gcc-help

Do I need the TREE_CODE there? Would TREE_TYPE (fld)) == METHOD_TYPE not
work?

On Wed, Jun 8, 2022 at 5:02 PM Xi Ruoyao <xry111@xry111.site> wrote:

> On Wed, 2022-06-08 at 14:47 +1000, Zopolis0 via Gcc-help wrote:
>
> /* snip */
>
> > java/lang/Object.java: internal compiler error: tree check: expected
> > function_decl, have field_decl in layout_class_method, at
> java/class.cc:2627
> >
> > I believe that this error is due to the following incorrect code at line
> > 2589 at gcc/java/class.cc:
> >   for (tree method_decl = TYPE_FIELDS (this_class);
> >
> > Before it was removed, this line was:
> >   for (method_decl = TYPE_METHODS (this_class);
> >
> > But I have since modified it (
> >
> https://github.com/Zopolis4/gcj/commit/1f38bc896a704290ca0b742c60c60a88d5e1fb07#diff-bf49cdc948b20b9f25afd3fbb36922a10e979a55ff9600b3b51d0ec54175b752L2592
> )
> > in accordance with 5aaa8fb40681ee66282d73dab8c8eccbf5ee0518
> >
> > Given that this was incorrect, what would be the correct way to replace
> > this instance of TYPE_METHODS? Were all my replacements of TYPE_METHODS
> in
> > gcc/java/class.cc wrong?
>
> 5aaa8fb is not a simple replacement.  For example:
>
>    /* If there are user-defined methods, they are deemed non-trivial.  */
> -  for (tmp = TYPE_METHODS (type); tmp; tmp = TREE_CHAIN (tmp))
> -    if (!DECL_ARTIFICIAL (tmp))
> +  for (tree fld = TYPE_FIELDS (type); fld; fld = DECL_CHAIN (fld))
> +    if (TREE_CODE (TREE_TYPE (fld)) == METHOD_TYPE && !DECL_ARTIFICIAL
> (fld))
>          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>        return true;
>
> I think you need to check if the field is really a method like this.
> --
> Xi Ruoyao <xry111@xry111.site>
> School of Aerospace Science and Technology, Xidian University
>

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2022-06-08  8:13 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-06-08  4:47 Incorrect replacement of TYPE_METHODS with TYPE_FIELDS causing segfault Zopolis0
2022-06-08  7:02 ` Xi Ruoyao
2022-06-08  8:13   ` Zopolis0

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).