Re: i2c tools build failure with gcc 4.6

[Guenter Roeck <linux-0h96xk9xTtrk1uMJSBkQmQ@public.gmane.org>]

On Thu, Dec 20, 2012 at 10:45:10AM +0100, Jean Delvare wrote:
> Hi Guenter,
> 
> On Wed, 19 Dec 2012 17:07:33 -0800, Guenter Roeck wrote:
> > Seen with latest i2c-tools from svn repository.
> 
> With the subversion repository having gone on vacation, "latest"
> doesn't mean much :(
> 
Oh, right, "the latest one I have", which is svn revision 6075. Is that the
latest ?

> > gcc 4.6:
> > 
> > gcc-4.6  -Llib -li2c -o tools/i2cdetect tools/i2cdetect.o tools/i2cbusses.o
> > tools/i2cdetect.o: In function `main':
> > i2cdetect.c:(.text.startup+0x557): undefined reference to `i2c_smbus_write_quick'
> > i2cdetect.c:(.text.startup+0x5e9): undefined reference to `i2c_smbus_read_byte'
> > collect2: ld returned 1 exit status
> > make: *** [tools/i2cdetect] Error 1
> > 
> > gcc-4.4:
> > gcc-4.4  -Llib -li2c -o tools/i2cdetect tools/i2cdetect.o tools/i2cbusses.o
> > 
> > [ no error ]
> > 
> > Any idea ?
> 
> I'm using gcc 4.6.2 (from openSUSE 12.1) and "the latest
> i2c-tools" (that would be from October 25th if I can trust the CHANGES
> timestamp) builds just fine for me. OTOH this is a linking problem, not
> building problem, so the version of "ld" may be more relevant. Are you
> using the same version of "ld" in both cases?
> 
> The error messages suggest that gcc (and subsequently ld) doesn't find
> the symbols exported by libi2c. I can't figure out any reason why it
> would be so. Can you show us the output of "nm lib/libi2c.so.0.1.0"?
> This is what I get here:
> 
> 0000000000201e30 a _DYNAMIC
> 0000000000201fe8 a _GLOBAL_OFFSET_TABLE_
>                  w _Jv_RegisterClasses
> 0000000000201e10 d __CTOR_END__
> 0000000000201e08 d __CTOR_LIST__
> 0000000000201e20 d __DTOR_END__
> 0000000000201e18 d __DTOR_LIST__
> 0000000000000f10 r __FRAME_END__
> 0000000000201e28 d __JCR_END__
> 0000000000201e28 d __JCR_LIST__
> 0000000000202028 a __bss_start
>                  w __cxa_finalize@@GLIBC_2.2.5
> 0000000000000c90 t __do_global_ctors_aux
> 0000000000000850 t __do_global_dtors_aux
> 0000000000202020 d __dso_handle
>                  U __errno_location@@GLIBC_2.2.5
>                  w __gmon_start__
> 0000000000202028 a _edata
> 0000000000202038 a _end
> 0000000000000cc8 t _fini
> 00000000000007c8 t _init
> 0000000000000cd8 t _real_fini
> 0000000000000830 t call_gmon_start
> 0000000000202028 b completed.5890
> 0000000000202030 b dtor_idx.5892
> 00000000000008d0 t frame_dummy
> 0000000000000900 T i2c_smbus_access
> 0000000000000c00 T i2c_smbus_block_process_call
> 0000000000000a40 T i2c_smbus_process_call
> 0000000000000a70 T i2c_smbus_read_block_data
> 0000000000000950 T i2c_smbus_read_byte
> 00000000000009a0 T i2c_smbus_read_byte_data
> 0000000000000b30 T i2c_smbus_read_i2c_block_data
> 00000000000009f0 T i2c_smbus_read_word_data
> 0000000000000ad0 T i2c_smbus_write_block_data
> 0000000000000980 T i2c_smbus_write_byte
> 00000000000009d0 T i2c_smbus_write_byte_data
> 0000000000000ba0 T i2c_smbus_write_i2c_block_data
> 0000000000000940 T i2c_smbus_write_quick
> 0000000000000a20 T i2c_smbus_write_word_data
>                  U ioctl@@GLIBC_2.2.5
> 
Hmm - I seem to be having the same.

> This could also be a Makefile dependency issue... I can't remember
> what, if anything, guarantees that the library is built before the
> tools which use it.
> 
Repeating the command doesn't help, though.

Maybe I am missing some commits ?

Guenter