On Feb 7, 9:16 am, Ronan Paixão <ronanpai...@yahoo.com.br> wrote:
<SNIP>
> I respect your opinions and actually I just wanted to post mine. Surely
> I don't have the skills even to start such a great venture as porting a
> zillion apps to work on Windows. Actually, I don't really intend it to
> use Cygwin or whatever wrapper there is, but to be a really native
> Windows app.
>
> I can't really complain, specially since I use Linux, but I'm posting my
> Point-of-view because I'm the only person I know in person (not counting
> net-friends) that uses Linux as the main OS. Yet, I'd like to recommend
> Sage to my friends and teachers, who all use Windows XP and Vista
> (without SUA). I believe that's where Sage could get a really good
> traction in the Windows world, since only technically inclined people
> venture into VM stuff (even if it's easy to do, just thinking about
> another OS in their machine makes some people frown).
Yes, we can all agree on that, but the port to native Windows is non-
trivial and a re-port to Cygwin and a port to SFU are intermediate
stepping stones.
Cygwin being a pain made me even consider not doing it at all again
(we certainly had that opinion about 6 months ago), but once things
got bogged down we thought it would be good enough for now.
SUA offers a whole set of advantages over Cygwin:
* the linker actually works for non-trivial number of DLLs
* it has a much better compiler than Cygwin
* It can produce 64 bit binaries using gcc
* one can use the SUA gcc as a front end to MSVC as a backend, i.e.
make the build system believe you compile your code with gcc, but in
actuality use MSVC. This does not create a complete native binary
since that MSVC version depends on a special libc, but at least we can
test how code works with MSVC without first having to sort out all the
build system issues
* the port is sponsored by MSR for whom running a system with SUA is
not a restriction
* MinGW does not offer a viable alternative since we have to fix
POSIX issues, too. Python 2.5 as well as 2.6 do not even seem to work
well with MinGW, so this is likely to be a giant source of pain. As
mentioned elsewhere: I want to port Sage and not fix toolchain issues
since I have other things to do :)
So if you take Cygwin, MinGW and SUA the last one wins hands down for
the task I am looking at. Given that first and foremost the job here
is to make MSR happy since they support the port it is my number one
priority. Having some sort of 100% doctest passing Sage port on
Windows is much better than no port at all even though only a minority
of Windows users can actually use it. The Cygwin port will also see
some fixes, but if I move heavily from SUA into the native port it
will likely suffer falling off the back of the truck since it is the
least pleasurable environment to work in. So someone else will have to
step up to help out. I am not saying it is you, but there are plenty
of devs out there using Cygwin and having the skillset to start
building Sage and fix some issues if it blows up.
> I don't really care which way you choose, being compiling with Cygwin,
> GCC, Intel compiler or MSVC, since the time you spend is yours and I
> appreciate anything in the direction of supporting Sage. I just would
> like, if possible, to be able to run it on "consumer" versions of
> Windows.
Absolutely, but wishing for a thing to happen and making it happen are
two very different pairs of shoes ;)
It is obvious that Sage on Windows available on anything up from 32
bit XP would be great, but given the current state of Cygwin I don't
see this in a very positive light. Note that MS provided patches for
binutils way back when SFU was the top dog, but the binutils
maintainer never integrated them. This might explain very much why the
ld on Cygwin is so pathetic compared to the SFU one, but this is a
theory for now. Either way, one giant reason why open source
mathematical software on Windows is in a less desirable state than on
POSIX system is plain and simply that Cygwin lacks polish.
> Thanks for hearing,
> Ronan
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
-~----------~----~----~----~------~----~------~--~---
