On Jul 13, 3:24 pm, "Dr. David Kirkby" <[EMAIL PROTECTED]>
wrote:
> On Jul 13, 8:00 pm, mabshoff <[EMAIL PROTECTED]> wrote:
Hi,
> > solaris express and it blew up on pari.
>
> > It really depends on the toolchain on Solaris and as is a gcc
> > configured with the Sun ld will not work for various reasons beyond
> > Pari.
>
> I've just built 4.3.1 with binutils 2.18. I will see how that works
> with Sage.
It won't. I got patches for trivial thing up to ATLAS, but ATLAS
itself needs a small binutils patch (or a patch to the ATLAS build
system which I am working on). Singular is also broken (more
specifically building libSingular, but I got a fix for that, too.
> > > Turns out
> > > the pari shared library is build without any PIC
> > > flag and that caused his linker to just give up.
> > > Turns out that the problem is in fact widespread
> > > across a number of archs where the linker emits
> > > a warning about DT_TEXTREL (usually) but still
> > > return a shared object. I know it affects me on
> > > linux-x86.
>
> > It does depend on the toolchain on Linux also, more about that below.
>
> If at all possible, this dependancy should be removed, or at least
> reduced. It's not ideal that one has to have very specific compiler
> versions, which use a particular linker and assembler. Given Sun
> distribute gcc, if possible it would be nice for Sage to build with
> that.
I current doubt that very much. Any gcc using the Sun ld will remain
broken for a long, long time until probably 20+ spkg have upstream
fixes not to use GNU ld flags per default.
> Of course, the Sun compiler suite would be preferable, as that is more
> fixed and you can just say to someone to use it. Experience tells me
> the goal posts don't move as much as they do with gcc.
>
> I've just built gcc-4.3.1 (the latest) from source code, but it is not
> a trivual process.
Yes, I agree there.
> Attempting to compile a recent gcc with an old gcc
> is likely to fail, as is trying to compile gcc with Sun's compiler.
> The only way I could build the latest gcc was to compile a later
> version than what I had, then use this to compile the latest version.
I usually build a C only 3.4.6 using the SFW binutils and move on from
there.
> I think I went
>
> 1) 4.0.2 (with Sun linker)
> 2) 4.0.3 (with latest gnu binutils)
> 3) 4.2.2 (with latest gnu binutils)
> 4) 4.3.1 (the latest gcc, with the latest binutils).
>
> If at all possible, this sort of thing should be avoided, as it makes
> it difficult for someone to build software.
>
> The fact the development tools on Solaris can be a bit of a mess is
> probably one reason that places like Blastwave and Sunfreeware don't
> tend to update gcc much.
>
> Dave
The plan is to provide a Sage toolchain for Solaris since everything
else is too broken and/or too variable to work out of the box. It is
the only way to build Sage reliably on Solaris and I do it on every
Solaris box I build Sage on.
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
