Re: [Gcl-devel] help fixing edit_double, failure to print 1d-6 for gcl-2.6.8

[David Billinghurst]

On 10/10/2013 12:47 PM, Camm Maguire wrote:
> David Billinghurst <address@hidden> writes:
>
> > That works.  Thanks.
> >
> > Happy to try and answer any mingw questions, but I am not an expert.
> > In fact, I have been spectacularly unsuccessful in hacking gcl on
> > windows -
> > all I have managed to do is keep gcl-2.6.8pre - and now 2.6.8 -
> > running with gcc-3.3.1 on Windows XP.
> Greetings!  I'm certainly no expert, but from my experience with mingw
> on a real windows machine, there definitely appears to be instabilities
> in the toolchain.  Utilities fail with 'permission denied', mysterious
> segfaults building gmp are not reproducible, etc.  It almost appears as
> if the entire mingw toolchain suffered from what gcl did until we moved
> all its memory up above a certain address range.
>
> First, I'd like to know if binaries built under some lowest common
> denominator windows will run on all others.  Windows people don't appear
> to build software frequently, as just one of these issues would grind
> Linux development to a complete halt.  Is it then fruitful to forget
> about chasing down the buildtime idiosyncrasies in all windows versions
> and focus for now on one build platform which will run on all others?
>
> Take care,
I build the maxima windows installer using gcl.  This binary appears to 
run on all Windows versions since WinXP - well over 5000 downloads per 
version and no real complaints.  This is built using the Mike Thomas 
recipe, with c1993-4 software. The current release is build with 
2.6.8pre 2013-01-01. gcl-2.6.8 with floating point printing patch 
appears to work works, although I haven't made a release using it.  I 
haven't tried 2.6.10pre recently.

I get close on XP using recent mingw/msys and gcc-4.[678] - some mixes 
work and some don't.  I haven't tried systematically, kept good records 
or reported my failures.  It would be good to get this working reliably.

I can't even run configure under Windows 7.  The failures appear 
mingw/msys related.  This may be a problem with the build machine, but I 
haven't tried to separate either of my kids from their laptops to try 
another machine.

Let me try some systematic tests over the next few days.