core2 not recognized

Tim Van Holder tim.vanholder at anubex.com
Sat Nov 14 09:53:03 CET 2009


Torbjorn Granlund wrote:
> "Jay DeMarrais" <jay at sandstorm.net> writes:
> 
>   But I ran:
>   
>   ./configure
>   
>   and it gave me:
>   
>   -pc' not recognizede 'core2.. Invalid configuration 'core2
>   -pc-cygwin failed /bin/sh ./confure.sub core2
>   
>   as the error message.  I even included the mispelling of 'recognized'
>   in the error message :)
>   
> Curious.  I grepped through the GMP sources, and the word recognizede
> does not occur anywhere.
> 
> Somebody else ran into the same on cygwin:
> http://gmplib.org/list-archives/gmp-bugs/2009-May/001470.html
> 
>   What else is there?  I also thought I followed the bug reporting
>   giving the version, etc. as well as this description.
>   
> Note that also "confure.sub" above is misspelled.
> 
> I cannot help debug this, since I don't run Windoze.
> I strongly suspect this is a Windoze bug or a Cygwin bug.

When you see strangely-formatted text like that on a Windows system,
it's usually because you're using tools that expect Unixy text files on
files that are DOS text (embedded carriage returns put the cursor back
at the start and the remainder of the line overtypes what was there).
If your cygwin is configured to use Unix-style text, try using a cygwin
tar or unzip to extract the gmp sources instead of using Winzip or the
like (since those can be inclined to switch things to DOS text).
Also make sure your PATH does not include a mix of cygwin and non-cygwin
tools (e.g. a mingw sed and a cygwin bash).

Alternatively, try running config.guess without arguments to see what it
reports. Then try running config.sub, passing it the name reported by
config.guess. It's likely one of those scripts is the cause of the
errors you get (if you also get it when passing an explicit triplet to
configure, then it's most likely config.sub).
You can also try again with the latest versions of these two files (they
get updated quite regularly), which you can get here:

http://git.savannah.gnu.org/cgit/config.git/tree/

Actually, I just read the mailing list archive link Torbjon provided and
it looks like it's config.guess that's going wrong:

Jim Skea at JimsNewLaptop ~/gmp-4.3.1
$ ./config.guess
-pc-cygwin

Jim Skea at JimsNewLaptop ~/gmp-4.3.1
$ ./configfsf.guess
i686-pc-cygwin

My guess is that it's config.guess reporting "i686<CR>-pc-cygwin",
causing the problem. Is there a particular reason GMP uses its own
config.guess (I assume configfsf.guess is the "official" one)?


More information about the gmp-bugs mailing list