[Bug c/50646] New: configure detects big endian on little endian system.

[Bug c/50646] New: configure detects big endian on little endian system.

[fzvqedi at v dot mintemail.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

             Bug #: 50646
           Summary: configure detects big endian on little endian system.
    Classification: Unclassified
           Product: gcc
           Version: 4.5.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: fzvqedi@v.mintemail.com


checking whether byte ordering is bigendian... yes

| int                                                                           
| main ()                                                                       
| {                                                                             
|                                                                               
|            /* Are we little or big endian?  From Harbison&Steele.  */         
|            union                                                              
|            {                                                                  
|              long int l;                                                      
|              char c[sizeof (long int)];                                       
|            } u;                                                               
|            u.l = 1;                                                           
|            return u.c[sizeof (long int) - 1] == 1;                            
|                                                                               
|   ;                                                                           
|   return 0;                                                                   
| }                                
                                                                               
                                                     configure:8171: result:
yes

that test will return 0 on _little endian_
obviously the code was copypasted from a function isBigEndian(), but the author
forgot that returning 0 in main means success.

the reason it came so far is that configure dumbly added some stuff like
#define ssize_t int
to auto-host.h which caused all previous checks to err, because the libc just
typedef'd those types, resulting in:
 error: two or more data types in declaration specifiers
or
error: 'long long long' is too long for GCC

[Bug c/50646] configure detects big endian on little endian system.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2011-10-07
                 CC|                            |rwild at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-07 08:39:39 UTC ---
So ... does this cause any problem for you in GCC?  [note: it's expanded from
AC_C_BIGENDIAN in gcc/configure.ac]

If not, it's a autoconf bug, not a GCC bug.

[Bug bootstrap/50646] configure detects big endian on little endian system.

[fzvqedi at v dot mintemail.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

--- Comment #2 from Roger Meyer <fzvqedi at v dot mintemail.com> 2011-10-07 20:50:23 UTC ---
indeed, it would cause problems. i.e. compile some stuff with wrong endianness
and later silently crash. 

in this specific case here the build stops later on on another error, caused by
another case of configure failing. 

but i may have well overlooked _this_ specific issue, fixing the others that
caused the build errors, and later being clueless why xgcc or something else
segfaults.

since autofocn is "built-in" into gcc, it's both a autofnoc / gcc bug,
and needs to be fixed here and there.

[Bug bootstrap/50646] configure detects big endian on little endian system.

[joseph at codesourcery dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

--- Comment #3 from joseph at codesourcery dot com <joseph at codesourcery dot com> 2011-10-07 21:46:20 UTC ---
Until comparatively recently, the only thing that cared about host 
endianness was decimal floating-point support.  However, now the lexer 
cares as well.

What is the host system for which the endianness test gets that far?  For 
most hosts the endianness would be determined based on the header tests.

[Bug bootstrap/50646] configure detects big endian on little endian system.

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

--- Comment #4 from Andreas Schwab <schwab@linux-m68k.org> 2011-10-07 22:27:41 UTC ---
Your bootstrap compiler must be broken, because the test is correctly
implemented: a successful run of the test program means little endian.  Since
the test exited with non-zero exit status, your system is identified as big
endian.

[Bug bootstrap/50646] configure detects big endian on little endian system.

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

--- Comment #5 from Andreas Schwab <schwab@linux-m68k.org> 2011-10-07 22:30:14 UTC ---
*** Bug 50647 has been marked as a duplicate of this bug. ***

[Bug bootstrap/50646] configure detects big endian on little endian system.

[fzvqedi at v dot mintemail.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

--- Comment #6 from Roger Meyer <fzvqedi at v dot mintemail.com> 2011-10-09 23:34:47 UTC ---
the host system is aboriginal linux 1.03
armv6l (little endian)
using gcc 4.2.1 on qemu
somehow configure seems to think that the used compiler is a crosscompiler, but
i have no idea why.

[Bug bootstrap/50646] configure detects big endian on little endian system.

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50646

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|---                         |INVALID

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
As reported, this is a bug in the original compiler that was being used.  This
has since been well tested in the 2 years since this was reported and no other
reports of this issue so closing.