On Apr 24, 12:19 am, "Dr. David Kirkby" <david.kir...@onetel.net>
wrote:
> mabshoff wrote:
<SNIP>
Hi David,
> > OK. Not that for gcc 4.2.2 the gfortran creates completely broken code
> > on Sparc, so the only toolchain I will be using is the one specified
> > above since it is well tested by me.
>
> But it's pretty much irrelavent if 4.2.2 creates buggy fortan
No, it isn't. It took me considerable time to determine that certain
doctest failures I saw were compiler issues where code was
miscompiled. Right now there is only one toolchain that I will support
when push comes to shove on Solaris and that is the one I provide.
Other people are free to use whatever they want, but the people who
pay me to work on Solaris want to use the latest stable gcc release,
i.e. I will switch to gcc 4.4.0 soon. Over time hopefully we can get
the SFW tools, other gcc releases and eventually Sun's tool suite to
work, but given that I have plenty of other things to fix at the
moment that just isn't high priority.
By the way: That toolchain I use on an US IIIi today blew up with an
internal compiler error on a T2000 (specifically gfortran). Since that
machine isn't on SkyNet I will see if gcc 4.3.3 or 4.4.0 fixes the
problem and otherwise open a ticket with the gcc folks. But it isn't a
high priority item since I have to fix a number of bugs before the 4.0
deadline and I don't have any time for detours :). The sparc binary I
produced works on that box, so we will be using that until further
notice.
> - I assume
> you have found at least one version which does not. But the point is,
> whatever method you use to create a tool chain could be scripted. I
> doubt that script would be that long either.
No it won't take long for a decent version. I have meant to so that
for a while, but to get it working reliably and only with the SFW
tools in $PATH will take some work. Since I will rebase my toolchain
on gcc 4.4.0 in the next couple days I will see what I can do.
In general I have posted 3.4.1 32 bit sparc and x86 tarballs that
include the toolchain and should just work on any Solaris 10 build at
http://www.sagemath.org/bin/solaris/index.html
There are also the stand alone toolchains there in case anyone wants
to build Sage, but note that the 3.4.1 binaries contain fixes not yet
upstream, so build at least on Sparc will fail while Sympow won't work
out of the box as is on x86. Let me know how they work in case anyone
does play with them. All known problems are listed at
http://wiki.sagemath.org/solaris
and the issues that do not have tickets yet will get them in the next
couple days.
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
