gcc/testsuite/gfortran.dg/pointer_intent_1.f90 failure on OSX

[FX Coudert]

> [karma] testsuite/gfortran.dg% gfc pointer_intent_1.f90
> [karma] testsuite/gfortran.dg% a.out
> Segmentation fault
>
> The test passes when the code is compiled with -O(1-3,s).

I can confirm that on i686-linux; it doesn't show up in the testsuite  
because it only fails at -O0, which is not tested by the dejagnu  
framework. Can you please file a bug?

> Thread 0 Crashed:
> 0   f951        0x0001bbd8 gfc_check_pointer_assign + 0x284 (expr.c: 
> 2406)
> 1   f951        0x00054e8c resolve_code + 0x3c8 (resolve.c:4995)
> 2   f951        0x00056450 resolve_codes + 0x28 (resolve.c:7321)
> 3   f951        0x000564c8 gfc_resolve + 0x2c (resolve.c:7349)
> 4   f951        0x00049ea0 gfc_parse_file + 0x138 (parse.c:3248)
> 5   f951        0x0006b028 gfc_be_parse_file + 0x14 (f95-lang.c:308)
> 6   f951        0x001047c8 toplev_main + 0xc38 (toplev.c:1054)
> 7   f951        0x00002a24 _start + 0x188 (crt.c:267)
> 8   dyld        0x8fe1a31c _dyld_start + 0x64

That doesn't match. You report a crash of the generated program,  
which I confirm, but you show the backtrace of a compiler (front-end)  
bug. I can't reproduce a segfault in the compiler on  
pointer_intent_1.f90, even with valgrind. The backtrace of the  
generated code is:

Program received signal SIGSEGV, Segmentation fault.
0x08048488 in nonpointer.1003 () at pointer_intent_1.f90:69
69           t%point = 7
Current language:  auto; currently fortran
(gdb) where
#0  0x08048488 in nonpointer.1003 () at pointer_intent_1.f90:69
#1  0x080482b8 in MAIN__ () at pointer_intent_1.f90:24
#2  0x080484b7 in main (argc=Cannot access memory at address 0x0
) at ../../../../trunk/libgfortran/fmain.c:18


FX