in reply to Installing Net::SSH::Perl - problems with Math::Pari

I tried installing Net::SSH::Perl on a Solaris 8/9 system and ran into numerous issues. After digging a bit, one of the README documents discussed building the required modules under Sun. I don't recall exactly which module mentioned this but I understood the problem was SUN Perl was built with Sun's C compiler. Trying to build under gcc had numberous problems because pieces were missing.

If you have Sun's C compiler it should work. I'm going from memory here but I believe if you switched from kpic to using fpic then it seemed to work. I didn't get very far with this and eventually switched to building under SLES8 Linux. This seemed to work very well. If this doesn't make sense let me know. I'm digging and can't find which module actually said that.

I also strongly recommend running "make test" after installing each module. It helped guide me a bit. There were failures during the build caused by missing dependancies. IMO installing this module was a little heavy on the dependancies however it works well once installed properly. I now have two servers running it just fine (under Linux).

I know this doesn't directly answer your question. I'm hoping it helps you find your way. Please let me know if you are still having problems. This was one of the toughest installations I've done with Perl.
  • Comment on Re: Installing Net::SSH::Perl - problems with Math::Pari

Replies are listed 'Best First'.
Re^2: Installing Net::SSH::Perl - problems with Math::Pari
by tweetiepooh (Hermit) on Feb 07, 2006 at 14:31 UTC
    Still no joy.

    I've installed Perl 5.8.8 built with gcc 3.4.2 on the box I'm trying to build Math::Pari on.

    I've tried to hide the GNU ld and as (use ccs versions) and visa versa.

    libpari-2.1.7 won't work at all, the beta 2.2.n does compile the library but fails to build Math::Pari. I get errors in Pari.xs

    function installep - initialization makes pointer from integer without a cast
    function my_gpui - prec undeclared
    function setprecision = prec undeclared
    function fillargvect = prec undeclared
    function XS_Math__Pari_interface?? = prec undeclared (replace ?? with digits)
    INIT_JMP_off undeclared
    INIT_SIG_off undeclared
    pari_modules undeclared
    helpmessages_highlevel undeclared

[reply]
      Haven't built Math::Pari for a while - last time I did build it I had good results following the instructions for the "manual procedure" that you'll find in the INSTALL file in the Math::Pari source tarball. (This way you don't have to actually build libpari.)

      I didn't run './Configure' as mentioned in step b), and I didn't bother with 'make test' as mentioned in step c).

      Have you tried that approach ? If so, at which step did you strike trouble ?

      Cheers,
      Rob
[reply]
        Sort of success.

        I finally got a build/test/install to work by using the portable code

        $ perl Makefile.PL machine=port

        [download]
        Obviously this is not ideal. If I follow the manual instructions then I get (note this is different from the initial problem as that used pari 2.2)
        First I need to edit the code that FTP's pari libs as I can't use exte
+rnal ftp.  All I did was point at localhost and put the archive in a 
+known place.

$ perl Makefile.PL
...
gets the code (2.1.7) and patches it.
id's my machine as sparcv9 so sets build to sparcv8-micro
...
$ make
...
(compiling gen1.c)
./pariinl.h: In function `gmul':
./pariinl.h:887: error: asm-specifier for variable `hiremainder' confl
+icts with asm clobber list
...

        [download]
        Now if I try to config and build libpari separately.
        $ make gp
...
(compiling mp.c)
../src/kernel/none/mp.c: In function `addrr':
../src/kernel/none/mp.c:598: error: asm-specifier for variable `overfl
+ow' conflicts with asm clobber list
../src/kernel/none/mp.c:634: confused by earlier errors, bailing out
...

        [download]
        I get the same errors if I manually get pari, unpack it to a sibling of the Math::Pari build area, Configure libpari then try to make Math::Pari.
[reply]
[d/l]
[select]

In Section Seekers of Perl Wisdom