For the records. Was: Internal compiler error ?

Porting a LFS/BLFS system built on a machine 1 to a machine 2 may result in a 
ported-system in 2 with a failing compiler. This seems to be due to a 
malfunctioning instruction MULX ( unsigned multiplication ) in the gmp 
procedure.

See Michael Shell contribution: 

https://www.mail-archive.com/[email protected]/msg02228.html

The correction was given for gmp-6.0.0a, the problem persists however in the 
actual gmp-6.1.0 of LFS.

As Michael Shell indicated, the problem can be fixed by replacing the code 
following the line 846 in config.guess:

        if (strcmp (modelstr, "haswell") == 0)
          {
            /* Some Haswells lack BMI2.  Let them appear as Sandybridges for
               now.  */
            CPUID (feature_string, 7);
            if ((feature_string[0 + 8 / 8] & (1 << (8 % 8))) == 0)
            modelstr = "sandybridge";
         }
by
        modelstr = "coreisbr";

and deleting in mpn/x86_64/fat/fat.c the line 314/315:

        if ((dummy_string[0 + 8 / 8] & (1 << (8 % 8))) != 0)
                CPUVEC_SETUP_coreihwl;

A general solution seems to be still pending

Edgar

-- 
Dr.-Ing. Edgar Alwers <[email protected]>
GPG Key ID:AD5C6F70
-- 
http://lists.linuxfromscratch.org/listinfo/blfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to