This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
c++/2541: Wild store in finish_builtin_type
- To: gcc-gnats at gcc dot gnu dot org
- Subject: c++/2541: Wild store in finish_builtin_type
- From: hpenner at de dot ibm dot bom
- Date: 12 Apr 2001 11:24:00 -0000
- Reply-To: hpenner at de dot ibm dot bom
>Number: 2541
>Category: c++
>Synopsis: Wild store in finish_builtin_type
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: unassigned
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Apr 12 04:26:01 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator: hpenner@de.ibm.bom
>Release: gcc-2.95.2
>Organization:
>Environment:
Hit on S390 64 bit, but problem also on any system
>Description:
The conversion from 0 to a pointer to member function das
crash when building the qt2 library. The reason is the following:
build_ptrmemfunc_type calls finish_builtin_type.
There, (decl2, line 2308) the statement
TYPE_STUB_DECL (type) = TYPE_NAME (type);
is not setting the common.chain to TYPE_NAME.
instead the first word of the lang_type is set to that
pointer. But the first 64 bit of the lang types contain
bitfields like 'is_signature'.
This bitfield is evaluated in 'build_c_cast'
the line (typeck.c line 5774)
if (IS_SIGNATURE (type)) is sometimes true in case
of 64 bit compiler (since is_signature is in the second 32 bit)
I don't understand enough of the frontend to fix that problem.
The line
>How-To-Repeat:
gcc tt.ii -O3.
look at lang_type field of type in 'build_c_cast'
>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: application/octet-stream; name="tt.ii"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="tt.ii"
Y2xhc3MgYW55IHsKcHVibGljOgoJaW50IGk7CiAgdm9pZCBhKGludCB4LGludCB5KXt9Owp9OwoK
Cgp0eXBlZGVmIHZvaWQgKGFueTo6KmFwKSAoaW50LGludCk7CgpzdGF0aWMgYXAgZj0wOwo=