[Bug c++/57367] New: Missing warning: array subscript is above array bounds

[Bug c++/57367] New: Missing warning: array subscript is above array bounds

[Gildos at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

            Bug ID: 57367
           Summary: Missing warning: array subscript is above array bounds
           Product: gcc
           Version: 4.6.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: Gildos at gmail dot com

I write this simple code in the file test.cpp:

#####
#include <stdio.h>

int main(int argc, char**argv)
{
    (void)argv;

    int pippo[100] ;
    pippo[1000] = 1 ;
    pippo[0] = pippo[0] ;

    return argc;
}
#####

I receive NO warning typing:
$ gcc -Wall test.cpp -o test

Thanks
Guido

[Bug c++/57367] Missing warning: array subscript is above array bounds

[glisse at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> ---
This is all dead code, gcc discards it before it looks at possible issues.

[Bug c++/57367] Missing warning: array subscript is above array bounds

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-05-22
     Ever confirmed|0                           |1
           Severity|normal                      |enhancement

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed as a frontend issue.  Nobody bothered to implement the trivial
cases there.

[Bug c++/57367] Missing warning: array subscript is above array bounds

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

--- Comment #3 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Yeah, shared between the front-ends, this isn't a C++-only issue, of course.

[Bug c++/57367] Missing warning: array subscript is above array bounds

[Gildos at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

--- Comment #4 from Guido Del Sarto <Gildos at gmail dot com> ---
I do not receive warnings with the following (no dead code).
#####
#include <stdio.h>

void warning( int pippo[100] )
{
    pippo[1000] += pippo[0];

    printf( "%d\n", pippo[1000] ) ;
}

int main(int argc, char**argv)
{
    (void)argv;

    int pippo[100] ;
    pippo[   0] = 0 ;
    pippo[1000] = 1 ;
    warning(pippo);

    return argc ;
}
#####

[Bug c++/57367] Missing warning: array subscript is above array bounds

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

--- Comment #5 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Remember that -O2 is required for this kind of middle-end check. A front-end
check would not.

[Bug c++/57367] Missing warning: array subscript is above array bounds

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu.org

--- Comment #6 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Disappointing, Richard, that in 4.8/4.9 (vs 4.7) we don't seem to warn at all
for the testcase in Comment#4 too. I'm wondering if the tree-vrp.c check
couldn't be done a bit earlier, I don't know.

[Bug c++/57367] Missing warning: array subscript is above array bounds

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57367

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Paolo Carlini from comment #6)
> Disappointing, Richard, that in 4.8/4.9 (vs 4.7) we don't seem to warn at
> all for the testcase in Comment#4 too. I'm wondering if the tree-vrp.c check
> couldn't be done a bit earlier, I don't know.

In 4.8/4.9 we optimize away the code and constant-propagate to the
printf (warning is inlined) before VRP.  -O2 -fno-inline warns.

I've repeatedly said that the proper place for static analysis is IPA
time before early optimizers are running.  You can trivially setup
constant/copy propagation without doing the actual transform there.