[Bug c/20327] New: -01 produce code which segfault on powerpc (linux/macOSx/aix)

[Martin dot Quinson at loria dot fr]

Hello,

the attached code, when compiled with -O1, produce a code which segfaults. I'm
quite embarrassed with the host/target/build triplet asked in the form, since I
tested it to segfault on several platforms, namly:
   powerpc-unknown-linux-gnu       gcc (GCC) 3.3.5 (Debian 1:3.3.5-8)     
   powerpc-ibm-aix5.1.0.0          gcc (GCC) 3.3.3                        
and also on Mac OSX (but I don't have all the version details from here since I
asked a friend to do the test for me -- I'll add them later if needed).

Interessingly enough, this test case does not fool the following platform:
   powerpc-ibm-aix5.2.0.0          gcc version = "2.9-aix51-020209"
But the bug I'm tracking in my full featured program also appear on this one, so
I suspect that my testcase is too reduced for this one. I can give you pointer
to the full featured archive, but I doubt it would help you. Whenever you solve
the issues shown by this test case, I'll produce a new one if needed :)

I usually compile my code with -g -O2 -g -Wall -Wunused  -Wmissing-prototypes
-Wmissing-declarations  -finline-functions  -Wshadow -Wpointer-arith
-Wchar-subscripts -Wcomment  -Wformat=2 -Wno-unused-variable
-Wno-unused-function -Wno-unused-label -Wwrite-strings  -Werror -ffast-math
-funroll-loops -fno-strict-aliasing 
on x86, and it didn't trigger anything. 

My project (and thus the test case) works on the following platforms (with gcc):
alphaev67-unknown-linux-gnu    hppa64-unknown-linux-gnu
i686-pc-linux-gnu              ia64-unknown-linux-gnu
m68k-unknown-linux-gnu         mipsel-unknown-linux-gnu
sparc64-unknown-linux-gnu      sparc-sun-solaris2.8
x86_64-unknown-linux-gnu

The code runs well in valgrind both with and without optimisation on x86, but
the poor valgrind-ppc gets completly fooled with the optimized ppc version.

I don't include the .i file since the printf is just there for cosmetics. It
also segfaults without it.

I'm not root on those machines, so installing a new gcc is somewhat impractical
for me (but not impossible). 

In the test case, changing the i value to a global helps a lot, but again, my
full featured project still shows the issue, even when the big structs are not
allocated on the stack. 

I'm a bit impressed to fill a bug against gcc, and I hope I do it right. If you
need any more information, just tell me.

Thanks for your devotion to this mandatory tool,
Mt.

-- 
           Summary: -01 produce code which segfault on powerpc
                    (linux/macOSx/aix)
           Product: gcc
           Version: 3.3.5
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: Martin dot Quinson at loria dot fr
                CC: Martin dot Quinson at loria dot fr,gcc-bugs at gcc dot
                    gnu dot org
 GCC build triplet: powerpc-unknown-linux-gnu
  GCC host triplet: powerpc-unknown-linux-gnu
GCC target triplet: powerpc-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20327