[Bug c/64363] New: Unresolved labels with -fcheck-pointer-bounds and -mmpx

[Bug c/64363] New: Unresolved labels with -fcheck-pointer-bounds and -mmpx

[christian.otterstad at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64363

            Bug ID: 64363
           Summary: Unresolved labels with -fcheck-pointer-bounds and
                    -mmpx
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: christian.otterstad at gmail dot com

I believe there is a problem with the -fcheck-pointer-bounds and -mmpx
arguments for GCC.

gcc test.c -c -std=gnu99 -O2 -Wall -fcheck-pointer-bounds -g -mmpx
/tmp/ccugNfJ8.s: Assembler messages:
/tmp/ccugNfJ8.s:67: Error: can't resolve `.L5' {*UND* section} - `.L6' {*UND*
section}

For the following source:


#include <stdio.h>

int main(int argc, char **argv)
{
        static int array = &&label_B - &&label_A;

        label_A:

        printf("%d\n", array);

        label_B:

        return 0;
}


This is derived from code found in vfprintf.c in glibc (after a lot of
rewriting). Attempting to compile glibc results in numerous such errors ("can't
resolve") of the same type. Omitting -fcheck-pointer-bounds and -mmpx allows it
(and all of glibc as well) to compile without errors.

Part of the following assembly generated (using -S) is given here:

__chkp_zero_bounds:
        .zero   16
        .data
        .align 4
        .type   array.2314, @object
        .size   array.2314, 4
array.2314:
        .long   .L5-.L6
        .section        .text.unlikely

No AS version appears to be able to assemble this, as .L5, .L6 doesn't exist
anywhere in the assembly source file.


At least the following versions appear to have this problem:

gcc (GCC) 5.0.0 20141211 (experimental) using GNU assembler (Gentoo 2.24 p1.4)
2.24
gcc (GCC) 5.0.0 20141215 (experimental) using GNU assembler (Gentoo 2.24 p1.4)
2.24
gcc (GCC) 5.0.0 20141216 (experimental) using GNU assembler (Gentoo 2.24 p1.4)
2.24
gcc (GCC) 5.0.0 20141216 (experimental) using GNU assembler (GNU Binutils)
2.25.51.20141215
gcc (GCC) 5.0.0 20141216 (experimental) using GNU assembler (GNU Binutils)
2.25.51.20141216
gcc (GCC) 5.0.0 20140925 (experimental) using GNU assembler (GNU Binutils)
2.24.51.20140422
gcc (GCC) 5.0.0 20141218 (experimental) using GNU assembler (Gentoo 2.24 p1.4)


Please note that gcc (GCC) 5.0.0 20140925 (experimental) is the same version
provided by Intel (in both source and binary form) to be used experimentally
with MPX.
https://software.intel.com/en-us/articles/intel-software-development-emulator
However I've also tried these other versions to see if the issue happened to be
resolved.

[Bug target/64363] Unresolved labels with -fcheck-pointer-bounds and -mmpx

[enkovich.gnu at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64363

Ilya Enkovich <enkovich.gnu at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |enkovich.gnu at gmail dot com

--- Comment #1 from Ilya Enkovich <enkovich.gnu at gmail dot com> ---
We copy function to instrument it but static var is initialized using labels
from the original function and thus we get unresolved links.  Suppose we would
get the same problem with non local gotos.

Suppose it would be more safe to just don't instrument such functions for now
and get back to it at the next stage 1.

[Bug target/64363] Unresolved labels with -fcheck-pointer-bounds and -mmpx

[ienkovich at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64363

--- Comment #2 from ienkovich at gcc dot gnu.org ---
Author: ienkovich
Date: Fri Jan 16 13:08:24 2015
New Revision: 219733

URL: https://gcc.gnu.org/viewcvs?rev=219733&root=gcc&view=rev
Log:
gcc/

    PR target/64363
    * ipa-chkp.h (chkp_instrumentable_p): New.
    * ipa-chkp.c: Include tree-inline.h.
    (chkp_instrumentable_p): New.
    (chkp_maybe_create_clone): Use chkp_instrumentable_p.
    Fix processing of not instrumentable functions.
    (chkp_versioning): Use chkp_instrumentable_p. Warn about
    not instrumentable functions.
    * tree-chkp.c (chkp_add_bounds_to_call_stmt): Use
    chkp_instrumentable_p.
    * tree-inline.h (copy_forbidden): New.
    * tree-inline.c (copy_forbidden): Not static anymore.

gcc/testsuite/

    PR target/64363
    * gcc.target/i386/chkp-label-address.c: New.


Added:
    trunk/gcc/testsuite/gcc.target/i386/chkp-label-address.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-chkp.c
    trunk/gcc/ipa-chkp.h
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-chkp.c
    trunk/gcc/tree-inline.c
    trunk/gcc/tree-inline.h

[Bug target/64363] Unresolved labels with -fcheck-pointer-bounds and -mmpx

[christian.otterstad at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64363

--- Comment #3 from Christian Otterstad <christian.otterstad at gmail dot com> ---
Great, it seems this corrected the issue, but a new problem that didn't appear
to exist before seems to have been introduced. I created a new bug issue for
it. Bug 64805: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64805

[Bug target/64363] Unresolved labels with -fcheck-pointer-bounds and -mmpx

[ienkovich at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64363

Ilya Enkovich <ienkovich at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |ienkovich at gcc dot gnu.org
         Resolution|---                         |FIXED

--- Comment #4 from Ilya Enkovich <ienkovich at gcc dot gnu.org> ---
Fixed