[Bug libfortran/54012] New: printf crush with -lgfortran

[Bug libfortran/54012] New: printf crush with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

             Bug #: 54012
           Summary: printf crush with -lgfortran
    Classification: Unclassified
           Product: gcc
           Version: 4.6.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: astroseger@gmail.com


It could be a bug of Magia distributive, not the gcc, but just in case.

The following code results in a segmentation fault if compile with -lgfortran! 

%%%%%%%%%%%%%%%%%

#include <stdio.h>
int main ()
{
   printf(
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
          "%% %% %% %% %% %% %% %% %% %%\n"
         );
}

%%%%%%%%%%%%%%%%%%
gcc test.c -lgfortran

(without -lgfortran would work fine)

Distributive: Mageia release 2 (Official) for x86_64

gcc -v 

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-mageia-linux-gnu/4.6.3/lto-wrapper
Target: x86_64-mageia-linux-gnu
Configured with: ../configure --prefix=/usr --libexecdir=/usr/lib
--with-slibdir=/lib64 --with-bugurl=http://bugs.mageia.org/
--mandir=/usr/share/man --infodir=/usr/share/info --enable-checking=release
--enable-languages=c,c++,ada,fortran,objc,obj-c++,java --enable-linker-build-id
--build=x86_64-mageia-linux-gnu --host=x86_64-mageia-linux-gnu
--with-cpu=generic --with-system-zlib --enable-threads=posix --enable-shared
--enable-objc-gc --enable-long-long --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-clocale=gnu --enable-java-awt=gtk
--with-java-home=/usr/lib/jvm/java-1.5.0-gcj-1.5.0.0/jre
--with-ecj-jar=/usr/share/java/eclipse-ecj.jar --enable-gtk-cairo
--disable-libjava-multilib --enable-ssp --disable-libssp --disable-werror
--with-ppl --with-cloog --with-python-dir=/lib/python2.7/site-packages
--enable-lto
Thread model: posix
gcc version 4.6.3 (GCC)

[Bug libfortran/54012] printf crush with -lgfortran

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

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

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> 2012-10-29 23:32:31 UTC ---
Can't reproduce with 4.6.3.

[Bug libfortran/54012] printf crush with -lgfortran

[kargl at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

kargl at gcc dot gnu.org changed:

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

--- Comment #2 from kargl at gcc dot gnu.org 2012-10-30 00:33:54 UTC ---
(In reply to comment #0)
> It could be a bug of Magia distributive, not the gcc, but just in case.
> 
> The following code results in a segmentation fault if compile with -lgfortran! 
> 
> %%%%%%%%%%%%%%%%%
> 
> #include <stdio.h>
> int main ()
> {
>    printf(
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>           "%% %% %% %% %% %% %% %% %% %%\n"
>          );
> }
> 
> %%%%%%%%%%%%%%%%%%
> gcc test.c -lgfortran
> 
> (without -lgfortran would work fine)

What does gdb tell you?

[Bug libfortran/54012] printf crash with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #3 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 09:52:04 UTC ---

> What does gdb tell you?


gdb tells:

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6

[Bug libfortran/54012] printf crash with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #4 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 09:56:49 UTC ---
(In reply to comment #1)
> Can't reproduce with 4.6.3.

Have you compiled with -lgfortran? If yes then probably this bug is some
specific problem of Mageia 2 distributive....

[Bug libfortran/54012] printf crash with -lgfortran

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #5 from Marek Polacek <mpolacek at gcc dot gnu.org> 2012-10-30 10:47:13 UTC ---
(In reply to comment #4)
> Have you compiled with -lgfortran?

Yeah.

[Bug libfortran/54012] printf crash with -lgfortran

[sgk at troutmask dot apl.washington.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #6 from Steve Kargl <sgk at troutmask dot apl.washington.edu> 2012-10-30 13:20:46 UTC ---
On Tue, Oct 30, 2012 at 09:52:04AM +0000, astroseger at gmail dot com wrote:
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012
> 
> --- Comment #3 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 09:52:04 UTC ---
> 
> > What does gdb tell you?
> 
> 
> gdb tells:
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6
> 

Not very helpful.  What does gdb tell you when you 
compiler with the -g option and issue bt at the gdb
prompt?

[Bug libfortran/54012] printf crash with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #7 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 14:03:27 UTC ---
(In reply to comment #6)
> On Tue, Oct 30, 2012 at 09:52:04AM +0000, astroseger at gmail dot com wrote:
> > 
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012
> > 
> > --- Comment #3 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 09:52:04 UTC ---
> > 
> > > What does gdb tell you?
> > 
> > 
> > gdb tells:
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6
> > 
> 
> Not very helpful.  What does gdb tell you when you 
> compiler with the -g option and issue bt at the gdb
> prompt?


(gdb) run
Starting program: /home/seger/TEMP/10/a.out

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6
(gdb) bt
#0  0x00007ffff779f8f2 in __parse_one_specmb ()
   from /lib64/libc.so.6
#1  0x00007ffff7781230 in vfprintf () from /lib64/libc.so.6
#2  0x00007ffff778b179 in printf () from /lib64/libc.so.6
#3  0x000000000040051a in main () at 1.c:4

[Bug libfortran/54012] printf crash with -lgfortran

[sgk at troutmask dot apl.washington.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #8 from Steve Kargl <sgk at troutmask dot apl.washington.edu> 2012-10-30 17:57:51 UTC ---
> > > > What does gdb tell you?
> > > 
> > > gdb tells:
> > > 
> > > Program received signal SIGSEGV, Segmentation fault.
> > > 0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6
> > > 
> > 
> > Not very helpful.  What does gdb tell you when you 
> > compiler with the -g option and issue bt at the gdb
> > prompt?
> 
> (gdb) run
> Starting program: /home/seger/TEMP/10/a.out
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x00007ffff779f8f2 in __parse_one_specmb () from /lib64/libc.so.6
> (gdb) bt
> #0  0x00007ffff779f8f2 in __parse_one_specmb ()
>    from /lib64/libc.so.6
> #1  0x00007ffff7781230 in vfprintf () from /lib64/libc.so.6
> #2  0x00007ffff778b179 in printf () from /lib64/libc.so.6
> #3  0x000000000040051a in main () at 1.c:4
> 

This does not look like a libgfortran issue.  Do you have
valgrind on your system?  Can you run the good and bad
executable under valgrind?  This appears to be a buffer
issue in your libc.

[Bug libfortran/54012] printf crash with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #9 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 19:35:15 UTC ---

> 
> This does not look like a libgfortran issue.  Do you have
> valgrind on your system?  Can you run the good and bad
> executable under valgrind?  This appears to be a buffer
> issue in your libc.

1. I have this bug on 3 different computers with installed magiea 2 x86_64 

2. Yes, it is very strange. It not look like libgfortran issue, but if I
compile without -l gfortran it works.

3. If I compile with -static it always works.

4. Just in case 
 "ldd good" looks like: 
        linux-vdso.so.1 =>  (0x00007fffe4f54000)
        libc.so.6 => /lib64/libc.so.6 (0x00007fc68961c000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fc6899a8000)
  "ldd bad":
        linux-vdso.so.1 =>  (0x00007fffd97ff000)
        libgfortran.so.3 => /usr/lib64/libgfortran.so.3 (0x00007f74e86ad000)
        libc.so.6 => /lib64/libc.so.6 (0x00007f74e8321000)
        libquadmath.so.0 => /usr/lib64/libquadmath.so.0 (0x00007f74e80ec000)
        libm.so.6 => /lib64/libm.so.6 (0x00007f74e7e6a000)
        /lib64/ld-linux-x86-64.so.2 (0x00007f74e89c3000)
so they use the same libc

5. valgrind with "bad" says:

==26497== Memcheck, a memory error detector
==26497== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
==26497== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
==26497== Command: ./a.out
==26497== 
==26497== Invalid read of size 1
==26497==    at 0x51A58F2: __parse_one_specmb (in /lib64/libc-2.14.1.so)
==26497==    by 0x518722F: vfprintf (in /lib64/libc-2.14.1.so)
==26497==    by 0x5191178: printf (in /lib64/libc-2.14.1.so)
==26497==    by 0x400519: main (in /home/seger/TEMP/a.out)
==26497==  Address 0x4023b3 is not stack'd, malloc'd or (recently) free'd
==26497== 
==26497== 
==26497== Process terminating with default action of signal 11 (SIGSEGV)
==26497==  Access not within mapped region at address 0x4023B3
==26497==    at 0x51A58F2: __parse_one_specmb (in /lib64/libc-2.14.1.so)
==26497==    by 0x518722F: vfprintf (in /lib64/libc-2.14.1.so)
==26497==    by 0x5191178: printf (in /lib64/libc-2.14.1.so)
==26497==    by 0x400519: main (in /home/seger/TEMP/a.out)
==26497==  If you believe this happened as a result of a stack
==26497==  overflow in your program's main thread (unlikely but
==26497==  possible), you can try to increase the size of the
==26497==  main thread stack using the --main-stacksize= flag.
==26497==  The main thread stack size used in this run was 16777216.
==26497==
==26497== HEAP SUMMARY:
==26497==     in use at exit: 3,769 bytes in 15 blocks
==26497==   total heap usage: 19 allocs, 4 frees, 11,909 bytes allocated
==26497==
==26497== LEAK SUMMARY:
==26497==    definitely lost: 0 bytes in 0 blocks
==26497==    indirectly lost: 0 bytes in 0 blocks
==26497==      possibly lost: 0 bytes in 0 blocks
==26497==    still reachable: 3,769 bytes in 15 blocks
==26497==         suppressed: 0 bytes in 0 blocks
==26497== Rerun with --leak-check=full to see details of leaked memory
==26497==
==26497== For counts of detected and suppressed errors, rerun with: -v
==26497== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)

6. valgrind with good says that all ok...

quite strange :)

[Bug libfortran/54012] printf crash with -lgfortran

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

Tobias Burnus <burnus at gcc dot gnu.org> changed:

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

--- Comment #10 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-10-30 22:04:33 UTC ---
(In reply to comment #9)
>   "ldd bad":
>         linux-vdso.so.1 =>  (0x00007fffd97ff000)
>         libquadmath.so.0 => /usr/lib64/libquadmath.so.0 (0x00007f74e80ec000)

Could you try linking libquadmath directly instead of libgfortran (which
implies libquadmath)?


The reason is that libquadmath installs, if possible, GLIBC's printf hooks, cf.
http://gcc.gnu.org/onlinedocs/libquadmath/quadmath_005fsnprintf.html
and the source code at
http://gcc.gnu.org/viewcvs/trunk/libquadmath/printf/quadmath-printf.c?view=markup

With those hooks, "printf("%Qe\n", float128_number)" will work. In principle,
it shouldn't cause those issues, but depending on the GLIBC version it might.


(If libquadmath is the cause, it also explains that it works with -static as
libgfortran only statically links libquadmath if needed.)

[Bug libfortran/54012] printf crash with -lgfortran

[astroseger at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

--- Comment #11 from Sergey Rodionov <astroseger at gmail dot com> 2012-10-30 22:37:40 UTC ---
(In reply to comment #10)
> (In reply to comment #9)
> >   "ldd bad":
> >         linux-vdso.so.1 =>  (0x00007fffd97ff000)
> >         libquadmath.so.0 => /usr/lib64/libquadmath.so.0 (0x00007f74e80ec000)
> 
> Could you try linking libquadmath directly instead of libgfortran (which
> implies libquadmath)?
> 
> 
> The reason is that libquadmath installs, if possible, GLIBC's printf hooks, cf.
> http://gcc.gnu.org/onlinedocs/libquadmath/quadmath_005fsnprintf.html
> and the source code at
> http://gcc.gnu.org/viewcvs/trunk/libquadmath/printf/quadmath-printf.c?view=markup
> 
> With those hooks, "printf("%Qe\n", float128_number)" will work. In principle,
> it shouldn't cause those issues, but depending on the GLIBC version it might.
> 
> 
> (If libquadmath is the cause, it also explains that it works with -static as
> libgfortran only statically links libquadmath if needed.)


Yes. you are right. The real problem is libquadmath. If I compile with
-lquadmath it crash... 

But interesting that quadmath_snprintf by itself works...

[Bug libquadmath/54012] printf crash with -lgfortran

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2013-06-23
          Component|libfortran                  |libquadmath
     Ever confirmed|0                           |1

--- Comment #12 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Per comment #11, this is a libquadmath issue. Note that it works for me
(without any valgrind error) from gcc version 4.6.4 up to trunk under
86_64-apple-darwin10 (without any valgrind error).

[Bug libquadmath/54012] printf crash with -lgfortran

[fxcoudert at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54012

Francois-Xavier Coudert <fxcoudert at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
                 CC|                            |fxcoudert at gcc dot gnu.org
         Resolution|---                         |WORKSFORME

--- Comment #13 from Francois-Xavier Coudert <fxcoudert at gcc dot gnu.org> ---
No feedback in the last 2 years, works for me on x86_64-linux at -m32 and -m64,
with all gcc versions from 4.6 to trunk. Closing.