GCC Bugzilla – Bug 8358
Ada accesses freed memory
Last modified: 2003-07-25 17:33:46 UTC
For more details see also the thread starting at: http://gcc.gnu.org/ml/gcc-bugs/2002-10/msg00764.html and especially these follow-ups: http://gcc.gnu.org/ml/gcc-bugs/2002-10/msg00828.html http://gcc.gnu.org/ml/gcc-bugs/2002-10/msg00859.html The problem is this error: $ ../../xgcc -B../../ -c -g -O2 -g -O2 -W -Wall -gnatpg -I. -I/cvs/gcc/gcc/ada s-intman.adb -v -save-temps Reading specs from ../../specs Configured with: /cvs/gcc/configure --prefix=/opt/gcc/3.3-devel --disable-nls --enable-threads=posix --enable-clocale=gnu --enable-__cxa_atexit --enable-shared --enable-languages=c,c++,f77,java,objc,ada,treelang --enable-checking=misc,tree,rtl,gc,rtlflag,gcac --with-system-zlib Thread model: posix gcc version 3.3 20021016 (experimental) ../../gnat1 -I. -I/cvs/gcc/gcc/ada -quiet -dumpbase s-intman.adb -g -g -gnatpg -O2 -O2 -W -Wall s-intman.adb -o s-intman.s +===========================GNAT BUG DETECTED==============================+ | 3.3 20021016 (experimental) (i686-pc-linux-gnu) Storage_Error stack overflow (or erroneous memory access)| | Error detected at s-intman.adb:113:20 | | Please submit bug report by email to gcc-bugs@gcc.gnu.org. | | Use a subject line meaningful to you and us to track the bug. | | Include the entire contents of this bug box in the report. | | Include the exact gcc or gnatmake command that you entered. | | Also include sources listed below in gnatchop format | | (concatenated together with no headers between files). | | (use plain ASCII or MIME attachment). | | See gnatinfo.txt for full info on procedure for submitting bugs. | +==========================================================================+ Please include these source files with error report compilation abandoned build_unit_elab (gnat_unit=1076058484, body_p=1, gnu_elab_list=0xa5a5a5a5) at /cvs/gcc/gcc/ada/trans.c:5324 5324 if (TREE_PURPOSE (gnu_elab_list) == NULL_TREE) (gdb) bt #0 build_unit_elab (gnat_unit=1076058484, body_p=1, gnu_elab_list=0xa5a5a5a5) at /cvs/gcc/gcc/ada/trans.c:5324 #1 0x4001d2d0 in ?? () #2 0x08078ecf in tree_transform (gnat_node=1355) at /cvs/gcc/gcc/ada/sinfo.h:813 #3 0x08071ee6 in gnat_to_code (gnat_node=1076025312) at /cvs/gcc/gcc/ada/trans.c:232 internal error - unimplemented function unk_lang_create_fundamental_type called. (gdb) p gnu_elab_list $1 = (union tree_node *) 0xa5a5a5a5 Release: GCC 3.3 CVS 20021016 Environment: i686-linux-gnu with glibc 2.2.5 How-To-Repeat: compile with --enable-checking=gcac and see it fail when building the gnat-tools (make -C gcc gnatlib_and_tools).
From: Graham Stott <graham.stott@btinternet.com> To: aj@suse.de Cc: gcc-gnats@gcc.gnu.org, gcc-patches@gcc.gnu.org Subject: Re: ada/8358: Ada accesses freed memory Date: Fri, 25 Oct 2002 19:59:05 +0100 Andreas, Here's a patch which I have bootstraped i686-linux-gnu-pc with all checking enabled. This patch introduces a new GC root which is used to prevent the pending eleaborations from being GC collected while they are being enumerated by build_unit_elab. Ok for mainline? Graham ada/ChangeLog * trans.c (gnu_pending_elaboration_lists): New GC root. (build_unit_elab): Use.. -------------------------------------------------------------------------- Index: trans.c =================================================================== RCS file: /cvs/gcc/gcc/gcc/ada/trans.c,v retrieving revision 1.21 diff -c -p -r1.21 trans.c *** trans.c 23 Oct 2002 07:33:34 -0000 1.21 --- trans.c 25 Oct 2002 18:52:50 -0000 *************** tree gnu_block_stack; *** 88,93 **** --- 88,97 ---- handler. Not used in the zero-cost case. */ static GTY(()) tree gnu_except_ptr_stack; + /* List of TREE_LIST nodes containing pending elaborations lists. + used to prevent the elaborations being reclaimed by GC. */ + static GTY(()) tree gnu_pending_elaboration_lists; + /* Map GNAT tree codes to GCC tree codes for simple expressions. */ static enum tree_code gnu_codes[Number_Node_Kinds]; *************** build_unit_elab (gnat_unit, body_p, gnu_ *** 5298,5303 **** --- 5302,5311 ---- if (gnu_elab_list == 0) return 1; + /* Prevent the elaboration list from being reclaimed by the GC. */ + gnu_pending_elaboration_lists = chainon (gnu_pending_elaboration_lists, + gnu_elab_list); + /* Set our file and line number to that of the object and set up the elaboration routine. */ gnu_decl = create_subprog_decl (create_concat_name (gnat_unit, *************** build_unit_elab (gnat_unit, body_p, gnu_ *** 5357,5362 **** --- 5365,5373 ---- poplevel (kept_level_p (), 1, 0); gnu_block_stack = TREE_CHAIN (gnu_block_stack); end_subprog_body (); + + /* We are finished with the elaboration list it can now be discarded. */ + gnu_pending_elaboration_lists = TREE_CHAIN (gnu_pending_elaboration_lists); /* If there were no insns, we don't need an elab routine. It would be nice to not output this one, but there's no good way to do that. */ ------------------------------------------------------------------------------
From: grahams@gcc.gnu.org To: gcc-gnats@gcc.gnu.org Cc: Subject: ada/8358 Date: 5 Nov 2002 00:49:45 -0000 CVSROOT: /cvs/gcc Module name: gcc Changes by: grahams@gcc.gnu.org 2002-11-04 16:49:45 Modified files: gcc/ada : ChangeLog trans.c Log message: PR ada/8358 * trans.c (gnu_pending_elaboration_lists): New GC root. (build_unit_elab): Use.. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ada/ChangeLog.diff?cvsroot=gcc&r1=1.265&r2=1.266 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ada/trans.c.diff?cvsroot=gcc&r1=1.21&r2=1.22
State-Changed-From-To: open->closed State-Changed-Why: Fix installed on mainline; I think this is sufficient for closing this bug even though it still might lurk in the GCC 3.2 tree. Thanks!